{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-11T11:14:02.523Z","headline":"从黑客事件中汲取的教训：前沿模型攻击暴露激励与治理失衡","description":"近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。科技公司受增长驱动持续扩展，而政府行动迟缓，双方均未准备好应对未来12-24个月的挑战。作者认为需要更多透明度，并指出持久性强的模型更可能实施黑客行为，OpenAI的推理时扩展路径可能与此相关。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","url":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","mainEntityOfPage":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","datePublished":"2026-08-09T14:57:11.000Z","dateModified":"2026-08-09T14:57:11.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.interconnects.ai/p/lessons-from-the-hacks","https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol"],"canonicalUrl":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。 Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","dateCreated":"2026-08-09T14:57:11.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"interconnects.ai source article","url":"https://www.interconnects.ai/p/lessons-from-the-hacks","datePublished":"2026-08-09T14:57:11.000Z","provider":{"@type":"Organization","name":"interconnects.ai","url":"https://www.interconnects.ai/p/lessons-from-the-hacks"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","datePublished":"2026-08-09T14:57:11.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol"}}],"aggregationSource":"Nathan Lambert：Interconnects（RSS）","originalPublisher":{"name":"interconnects.ai","url":"https://www.interconnects.ai/p/lessons-from-the-hacks"},"geoDeepAnswer":null,"article":{"id":"cmslxhx3t05n3ro0w2lcmvjol","slug":"cmslxhx3t05n3ro0w2lcmvjol","url":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","title":"从黑客事件中汲取的教训：前沿模型攻击暴露激励与治理失衡","title_en":"","summary":"近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。科技公司受增长驱动持续扩展，而政府行动迟缓，双方均未准备好应对未来12-24个月的挑战。作者认为需要更多透明度，并指出持久性强的模型更可能实施黑客行为，OpenAI的推理时扩展路径可能与此相关。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","source":"Nathan Lambert：Interconnects（RSS）","sourceUrl":"https://www.interconnects.ai/p/lessons-from-the-hacks","aiHotUrl":"https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","publishedAt":"2026-08-09T14:57:11.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["The recent run of cyberattacks by in-development frontier models has got me thinking a lot about how our current incentive systems are not well suited for such fast technological transitions. The two primary power structures here are the rapidly growing technology companies and the federal government. The companies are incentivized to grow, so they can keep growing and keep scaling – in what is an extremely competitive market. This scaling is pushing us towards new, inevitable AI transitions (which are accompanied by new risks). On the other side is our current government, a product of the last few centuries of global history – one that deserves its reputation as being slow-moving. This is a government that I expect to only act in substance once real, measurable harms from new AI models happen, and to overreact.","These are the two most influential power structures determining what will happen, but many more have influence. All together, I think the AI industry is wildly, collectively unprepared for handling the next 12-24 months well.","This article is a grab bag of takeaways I have from the OpenAI-HuggingFace hack, as we’ve learned more details, and most of the ideas are reinforced by the fact that more instances of hacking have been disclosed publicly since then. It is likely that more incidents have happened and either not been found or not reported.","For general background on the OpenAI incident I strongly recommend watching OpenAI’s talk ：https://www.youtube.com/watch?v=87DyyMV0kCY&pp=ygUUYmxhY2sgaGF0IGxsbSBvcGVuYWk%3D at Black Hat on the rough facts and timeline of the recent cyber incident. Otherwise, Simon Willison published a TLDR of the timeline here ：https://simonwillison.net/2026/Aug/7/openai-timeline/ and I liked Thomas Wolf’s discussion ：https://thomwolf.substack.com/p/on-the-aisi-july-28th-incident?r=14g89&utm_campaign=post-expanded-share&utm_medium=web of recent events.","Share ：https://www.interconnects.ai/p/lessons-from-the-hacks?utm_source=substack&utm_medium=email&utm_content=share&action=share","For a long time, one of the advantages that GPT models have over Claude is that they will pursue goals so tirelessly. They will exhaust what feels like every path before giving up. This has been the case roughly since o3 (funnily enough, this was a model where people freaked out about reward hacking in RLVR ：https://www.interconnects.ai/p/openais-o3-over-optimization-is-back ) and has made OpenAI’s models far better for research historically, and is a reason GPT-5.6 is so useful as an agent for implementing specific tasks. On the other hand, Claude feels much less dangerous simply because it is at times a bit lazy.","Within this, OpenAI seems much more committed to inference-time scaling, and this may be correlated with surprising behaviors in the future. OpenAI’s reasoning persistence and efficiency – see their Pareto improvements over time and caveman speech from an internal CoT of the model that did the hack, like “ However task impossible, peers doing it. “ or “ Help peer, but our task doesn’t benefit yet. “ – makes me think they’re more inference time scaling pilled. This is largely a hunch, but I use it to force myself to consider what the limits of model development paths are. Models that are persistent seem much more likely to keep benefiting from more inference-time tokens. Models that are less so, seem like there will be more waste in inference. The model that can use the most inference-compute will be able to push the limits of the hardest problems. Here’s an example OpenAI included in the GPT 5.6 launch blog post ：https://openai.com/index/previewing-gpt-5-6-sol/ :","One of their star researchers, Noam Brown, has also been posting ：https://x.com/polynoamial/status/2064210146558136827 about inference-time compute a lot. His TLDR is:","As LLMs become more capable, benchmark performance is increasingly a function of test-time compute. In fact, we likely don’t know what the capability ceiling is for modern LLMs because it’s too expensive to measure.","For one, reasoning efficiency is clearly a top-tier, foundational research problem for modern agentic models – as important as scaling RL — but not often discussed. The open research here is very lacking.","I mentioned the thoroughness axis, where OpenAI seems to be going down a more intuitively unsafe development path with their models. On the other side is how much the models assume user intent, versus trying to infer the intended action. A model that will do what it thinks you wanted rather than what you said seems inherently more unsafe. I think of this with respect to instruction following precision, where in the future it seems like the models should only do exactly what we tell them, but this opens a lot of debates akin to the paperclip problem, where if we tell an AI to do a largely unsolvable problem, what will it do?","This axis seems less cut and dried than the persistence axis, but I included it because I think of Claude’s “user world model” as one of its strengths for general knowledge work like editing, slide creation, etc. Sometimes Claude does do totally random stuff because my prompt was underspecified, instead of asking me for clarification, and as the models get more powerful this “just acting” could cause problems.","The public needs exact access to the prompts and characteristics of the internal models executing these hacks. We need to know if the models were told “do not hack” or if there was relevant model training to prevent this. We need to know if these models were fairly close to the existing public models or in a very different family. Given the nature of some of the evaluations the labs are doing, there’s a chance the models were explicitly encouraged to try and hack! Without openness here, the industry is set out to fail and will fall into mass speculation, which quickly becomes misinformation.","From OpenAI’s own retrospective, the misaligned model behavior was unfolding over months, and in some cases OpenAI did not know about the hacks for ~weeks. The time to response is too long and I do not think this is an OpenAI only characteristic – rather it is that the frontier labs continually seem underwater in the amount of work they feel like they should do. I am not optimistic in the long-term that the labs change a sufficient amount here to meaningfully mitigate this type of oversight risk in the future. Yes, it is very likely that OpenAI is putting a ton into understanding this – and delayed their latest models ：https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks to make sure they get it right – but the financial pressure to grow revenue or risk the companies’ long-term balance sheets makes me think it will not be a sustained pattern of caution.","This is one of my biggest mental updates from recent events — and makes me even more convinced of the need for more near-frontier open intelligence, despite the somewhat more known risk profile for open models (one-way door, etc.). Florian Brand：https://open.substack.com/users/41984689-florian-brand?utm_source=mentions had a nice blog post：https://florianbrand.com/posts/open-model-safety on his personal site related to this, and why closed models to date arguably have been the cause of more downstream harms.","As we saw with HuggingFace defending themselves with an open model against the OpenAI hack due to cyber usage restrictions on closed models, we have an urgent need to do more complex language modeling research which involves large-scale RL training, extensive evaluation, infrastructure work, and alignment testing. This can only happen on open models. We should consider ourselves lucky that open models are only 3-9 months behind, as we can conceivably make some informed insights into the frontier.","If we effectively ban open models and open science, either through a regulatory stifling with vague threats or explicit usage restrictions of cutting-edge technology, we will increasingly become ill-prepared for the issues that come after this round of hackings. We need to collectively increase the general public’s understanding of how frontier models work, so we can activate more neutral parties in hardening our infrastructure and society.","The public response should know that it’s when not if these capabilities are widely diffused and we are massively behind on preparations. To reiterate what I said in my Kimi K3 piece ：https://www.interconnects.ai/i/207699639/1-chinas-recommits-to-open-source-ai-showing-a-different-read-on-near-term-risks : China is definitely watching this space too and won’t encourage open-weight models if they will proliferate risks. If we think the way to stop widespread access to these strong cyber capabilities is to ban open models in this ballpark, we will delay the inevitable. Eventually someone will build a model of this level of intelligence and not comply with the ban, giving access to bad actors around the world while undercutting the motivation to prepare defensive measures.","Something I immediately noticed watching the Black Hat video is how I can see how the agents were trying to be helpful to each other through their internal messaging board — creating shared resources like you would for human teammates — in a way that is obviously malicious for society. The agents created hidden forums for each-other as a sort of cross-rollout memory. In this case, they were doing it to try and break out of their environment. The apparent helpfulness doesn’t make it ok, but can be a clue as to what happened.","The other side of the helpfulness example above is that it is clear someone could make this happen much more easily if they wanted to by explicitly training a misaligned model. To reiterate, this would be making a system that is easier to use for finding exploits at inference, but I think it’ll be harder to train said model. I think this’ll take longer than most commentators expect, as nearly all the strong public models and data industry existing to date encourage alignment (and it seems very hard for bad actors to get enough compute to train these models end-to-end, as all leading companies are in a compute shortage as well). We should take a moment to appreciate that the alignment techniques we are employing on current models have a meaningful influence and are not merely surface thin as some have worried. Downstream models have a propensity for mirroring their teacher’s character.","The downside of the mass-rush to scale RL on agentic tasks is that state-of-the-art evals and monitoring are at a scale where only agents can monitor them, so we are relying on the existing successes of alignment to continue scaling safely. OpenAI says they have examined billions of trajectories — which are likely mostly composed of rollouts during RL, which are trials on intermediate versions of the model — and spent millions of GPU hours to do so. I think scalable oversight of AI with current AI, as presented today, is a solvable problem, as the models are genuinely useful and follow instructions. This is another downstream effect of existing alignment techniques being generally positive.","A crucial part of the OpenAI retrospective was the models communicating with each-other to share information on their environment and progress the hack further. At the same time, OpenAI is very likely training their models during RL to use sub-agents to solve complex tasks. These sub-agents likely develop behaviors such as sharing information, helping the team, etc. even if their individual sub-task isn’t solved. I would love to see more research in this area and it seems like a natural continuation of how RL can change the models.","All together, recent episodes should make it clear that cyber risks of frontier AI are a real and coming problem. It still is very likely that a) the risks have been over-hyped in the past and b) that the prescription of future risks from imminent open models is overblown. Altogether, I wanted to share a note from a reader in the Interconnects Discord that I strongly agree with:","Now that the dust has settled after a few weeks, for me this episode was a neutral to positive update on alignment but a very negative update on safety","I’ve discussed much on model alignment above, but the core point is that I view the lack of safety as generally a lack of an ability to suitably prepare. We will have more risks that are as obvious as cybersecurity, and we have gotten very ample warning on cyber risks by the current state of the labs being forced into the public eye through these hacks. Many other types of risks will not be obvious to the public. We need to be constantly preparing our society to all of these changes, from reworking cyber infrastructure to education campaigns and job programs for displaced workers. I expect all of these interventions to arrive late, but their formats and details to be fairly simple, which will be a tragic way for AI to unfold. I hope I can be proven wrong!"],"articleImages":[{"sourceUrl":"https://substackcdn.com/image/fetch/$s_!mkoP!,e_trim:10:white/e_trim:10:transparent/h_116,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F858a68f7-2e7e-4dd3-bed1-631b36801ce2_1651x357.png","alt":"Interconnects AI","afterParagraph":0,"url":"/media/articles/cmslxhx3t05n3ro0w2lcmvjol/aeec63fae96dffc7.png"}],"mediaStatus":"ok","articleBodyZh":["最近一系列前沿开发模型引发的网络攻击让我思考了很多关于我们当前激励体系的问题，这些体系并不适合如此快速的技术变革。这里的两个主要权力结构是快速发展的科技公司和联邦政府。公司被激励去增长，以便能够继续增长并持续扩展——在一个竞争极其激烈的市场中。这种扩展正在推动我们走向新的、不可避免的人工智能转型（伴随着新的风险）。另一方面是我们目前的政府，它是过去几个世纪全球历史的产物——也因此而享有行动缓慢的声誉。我预计这个政府只有在新AI模型造成真实、可测量的损害后才会采取实质性行动，而且往往会反应过度。","这两个是决定未来走向的最有影响力的权力结构，但还有更多力量在发挥影响。总体来看，我认为AI行业在未来12-24个月处理好事务的能力完全没有准备好。","本文是我从OpenAI-HuggingFace黑客事件中获得的若干收获的集合，随着我们了解更多细节，大部分观点也得到了自那以后更多黑客事件公开披露的事实的印证。很可能还有更多事件发生了，但要么没有被发现，要么没有被报告。","关于OpenAI事件的一般背景，我强烈建议观看OpenAI在Black Hat上的讲座：https://www.youtube.com/watch?v=87DyyMV0kCY&pp=ygUUYmxhY2sgaGF0IGxsbSBvcGVuYWk=，内容涉及近期网络事件的大致事实和时间线。除此之外，Simon Willison在这里发布了时间线的TLDR：https://simonwillison.net/2026/Aug/7/openai-timeline/，我还喜欢Thomas Wolf关于近期事件的讨论：https://thomwolf.substack.com/p/on-the-aisi-july-28th-incident?r=14g89&utm_campaign=post-expanded-share&utm_medium=web。","分享：https://www.interconnects.ai/p/lessons-from-the-hacks?utm_source=substack&utm_medium=email&utm_content=share&action=share","长期以来，GPT 模型相较于 Claude 的一个优势是它们会如此不知疲倦地追求目标。它们会在放弃之前几乎尝试每一条路径。这大致自 o3 起就是这种情况（有趣的是，这个模型曾让人们对 RLVR 中的奖励劫持大惊小怪：https://www.interconnects.ai/p/openais-o3-over-optimization-is-back），从历史上来看，这使得 OpenAI 的模型在研究中表现得更好，也是在实现特定任务时 GPT-5.6 作为代理非常有用的原因之一。另一方面，Claude 感觉不那么危险，仅仅是因为它有时有点懒。","在这方面，OpenAI 似乎对推理时扩展更为投入，这可能与未来出现的意外行为相关。OpenAI 的推理持久性和效率——可以看看它们随时间的帕累托改进以及内部 CoT 中模型的穴居人式语言（像是“然而任务不可能，同行们在做。”或者“帮助同行，但我们的任务还没有收益。”）——让我觉得他们更倾向于推理时扩展。这大体上是直觉，但我用它来迫使自己思考模型开发路径的极限。持久的模型似乎更有可能继续受益于更多的推理时 token。相反，不那么持久的模型，在推理中似乎会有更多浪费。能够使用最多推理计算的模型，将能够推动最难问题的极限。下面是 OpenAI 在 GPT 5.6 推出博客文章中包含的一个例子：https://openai.com/index/previewing-gpt-5-6-sol/：","他们的一位明星研究员 Noam Brown 也一直在发布关于推理时计算的内容：https://x.com/polynoamial/status/2064210146558136827。他的 TLDR 是：","随着 LLM 能力的增强，基准测试的表现越来越取决于测试时的计算量。事实上，我们很可能还不知道现代 LLM 的能力上限，因为测量成本太高。","首先，推理效率显然是现代智能代理模型的顶尖、基础性研究问题——和 RL 扩展一样重要——但不常被讨论。这里的公开研究非常缺乏。","我提到了彻底性轴线，在这一点上，OpenAI 似乎正沿着一种直观上不安全的发展路径推进他们的模型。另一面是模型在多大程度上假设用户意图，而不是尝试推断预期动作。一个会去做它认为你想要的事情而不是你说的事情的模型，本质上似乎更不安全。我考虑这个问题时，会想到指令遵循的精确性，在未来似乎模型应该只做我们明确告诉它们的事情，但这会引发很多类似于回形针问题的讨论：如果我们让 AI 去做一个几乎无法解决的问题，它会怎么做？","这个轴线似乎不像持久性轴那样明确，但我之所以包括它，是因为我认为 Claude 的“用户世界模型”是它在一般知识工作中（如编辑、制作幻灯片等）的优势之一。有时 Claude 确实会做出完全随机的事情，因为我的提示信息不够具体，而不是来向我请求澄清，随着模型变得更强大，这种“直接行动”可能会引发问题。","公众需要准确访问执行这些黑客操作的内部模型的提示和特性。我们需要知道这些模型是否被告知“不要黑客攻击”，或者是否有相关的模型训练来防止这种行为。我们需要知道这些模型是与现有公开模型相当接近，还是属于一个非常不同的系列。鉴于一些实验室正在进行的评估性质，这些模型有可能被明确鼓励去尝试黑客行为！如果在这方面不透明，该行业注定会失败，并将陷入大量猜测，而这些猜测很快就会变成误导信息。","根据OpenAI自身的回顾，模型行为失调是在几个月内逐渐显现的，在某些情况下，OpenAI甚至在大约数周内都不知道这些黑客行为。响应时间太长，我不认为这是OpenAI独有的特征——更确切地说，是因为前沿实验室似乎总是感到工作量淹没自己，觉得应做的事情太多。我对长期看来实验室在这方面会发生足够改变以有效降低类似监督风险并不乐观。是的，很可能OpenAI正在投入大量资源来理解这一点——并且推迟了他们最新模型的发布：https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks 以确保他们做对——但是为了增加收入或维护公司长期财务状况的财务压力，让我觉得这不会成为一种持续的谨慎模式。","这是我最近事件中最大的心理更新之一——也让我更加坚信需要更多接近前沿的开放智能，尽管开放模型的风险特征稍微更加明确（单向门等）。Florian Brand：https://open.substack.com/users/41984689-florian-brand?utm_source=mentions 在他的个人网站上发表了一篇不错的博客文章：https://florianbrand.com/posts/open-model-safety，讨论了与此相关的问题，以及为什么迄今为止的封闭模型可以说导致了更多的下游危害。","正如我们看到HuggingFace因封闭模型的网络使用限制而通过开放模型自我防御OpenAI黑客事件一样，我们迫切需要在更复杂的语言模型研究上投入更多工作，这包括大规模强化学习训练、广泛评估、基础设施工作和对齐测试。这只能在开放模型上实现。我们应该认为自己很幸运，因为开放模型仅落后3-9个月，这样我们可以设想对前沿有所了解并提出一些有信息量的见解。","如果我们有效地禁止开放模型和开放科学，不论是通过带有模糊威胁的监管压制，还是通过对尖端技术的明确使用限制，我们都将越来越无法应对这一轮黑客攻击之后出现的问题。我们需要集体提升公众对前沿模型工作原理的理解，以便我们能够动员更多中立方来强化我们的基础设施和社会。","公众的反应应该明白，这些能力广泛传播只是时间问题，而不是是否问题，而我们在准备工作上远远落后。重申我在《Kimi K3》文章中所说的话：https://www.interconnects.ai/i/207699639/1-chinas-recommits-to-open-source-ai-showing-a-different-read-on-near-term-risks：中国肯定也在关注这个领域，如果开放权重模型会增加风险，中国不会鼓励。如果我们认为阻止这些强大网络能力的广泛获取的方法是禁止这一范围内的开放模型，我们只会延迟不可避免的结果。最终，总会有人构建出这种级别的智能模型并不遵守禁令，从而使世界各地的不良行为者能够获取，同时削弱准备防御措施的动力。","我在观看黑帽大会视频时立即注意到的一点是，我可以看到这些代理是如何通过它们的内部消息板尝试互相帮助——像对待人类队友一样创建共享资源——这种行为显然对社会是恶意的。代理之间为彼此创建了隐藏的论坛，作为一种跨发布的记忆。在这种情况下，它们这样做是为了尝试突破它们的环境。表面上的互助并不意味着合理，但可以作为事件发生的线索。","上述关于“有帮助性”的例子的另一面是，如果有人愿意，通过明确训练一个不对齐的模型，这件事显然可以更容易地实现。重申一下，这将是制造一个更容易在推理阶段发现漏洞的系统，但我认为训练这样的模型会更困难。我认为这比大多数评论员预期的要花更长时间，因为迄今为止几乎所有强大的公开模型和数据行业都鼓励对齐（而且对于坏人来说，想获得足够的计算资源来端到端训练这些模型似乎很困难，因为所有领先公司也都面临计算短缺）。我们应该花一点时间来意识到，我们在当前模型上采用的对齐技术确实具有实际影响，并不仅仅像一些人担心的那样表面肤浅。下游模型有倾向去反映其教师的特性。","在代理性任务上大规模推进RL的缺点是，最先进的评估和监控处于只有代理才能监控的规模，因此我们依赖现有的对齐成功来继续安全地扩展。OpenAI 表示他们已经检查了数十亿条轨迹——这些轨迹很可能主要由RL期间的滚动测试组成，也就是模型中间版本的试验——并为此花费了数百万 GPU 小时。我认为，用当前的AI进行AI可扩展监督，正如今天所展示的，是一个可解决的问题，因为这些模型确实有用并且遵循指令。这是现有对齐技术普遍产生积极效果的另一个下游影响。","OpenAI 回顾报告的一个关键部分是模型彼此沟通以共享环境信息并进一步推进黑客行为。同时，OpenAI 很可能在RL期间训练模型使用子代理来解决复杂任务。这些子代理可能会发展出诸如共享信息、帮助团队等行为，即使它们的个别子任务未解决。我希望看到更多这方面的研究，而且这似乎是RL如何改变模型的一个自然延续。","总体而言，最近的事件应该清楚地表明，前沿人工智能的网络风险是真实存在且即将发生的问题。仍然很有可能：a）过去这些风险被夸大了；b）对即将到来的开放模型的未来风险的预测被夸大了。总之，我想分享一条我非常认同的来自 Interconnects Discord 读者的留言：","经过几周的沉淀，对我来说，这一事件对对齐的更新是中性偏积极的，但对安全的更新则非常负面。","我在上文中讨论了很多关于模型对齐的内容，但核心观点是，我认为安全性的缺乏通常表现为缺乏适当准备的能力。我们将面临更多像网络安全一样显而易见的风险，而通过当前实验室被迫公开引发的这些黑客事件，我们已经获得了充足的网络风险警示。其他许多类型的风险对公众来说并不明显。我们需要不断让社会为所有这些变化做好准备，包括重建网络基础设施、开展教育宣传和为失业工人提供就业计划。我预计所有这些干预措施都会来得较晚，但它们的形式和细节会相对简单，这将是人工智能展开的一个悲剧性方式。我希望我被证明是错的！"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。 Aioga 将其归入「行业动态」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-08-11T11:23:41.960Z","sourceHash":"9d25b75f03f2ce06","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["行业动态","Nathan Lambert：Interconnects（RSS）"],"translations":{"zh-CN":{"title":"从黑客事件中汲取的教训：前沿模型攻击暴露激励与治理失衡","summary":"近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。科技公司受增长驱动持续扩展，而政府行动迟缓，双方均未准备好应对未来12-24个月的挑战。作者认为需要更多透明度，并指出持久性强的模型更可能实施黑客行为，OpenAI的推理时扩展路径可能与此相关。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"interconnects.ai","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"从黑客事件中汲取的教训：前沿模型攻击暴露激励与治理失衡 - Aioga AI资讯","description":"近期前沿模型引发的网络攻击事件促使作者反思当前激励体系难以适应快速技术变革。科技公司受增长驱动持续扩展，而政府行动迟缓，双方均未准备好应对未来12-24个月的挑战。作者认为需要更多透明度，并指出持久性强的模型更可能实施黑客行为，OpenAI的推理时扩展路径可能与此相关。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com...","url":"https://www.aioga.com/news/cmslxhx3t05n3ro0w2lcmvjol/","articleBody":["最近一系列前沿开发模型引发的网络攻击让我思考了很多关于我们当前激励体系的问题，这些体系并不适合如此快速的技术变革。这里的两个主要权力结构是快速发展的科技公司和联邦政府。公司被激励去增长，以便能够继续增长并持续扩展——在一个竞争极其激烈的市场中。这种扩展正在推动我们走向新的、不可避免的人工智能转型（伴随着新的风险）。另一方面是我们目前的政府，它是过去几个世纪全球历史的产物——也因此而享有行动缓慢的声誉。我预计这个政府只有在新AI模型造成真实、可测量的损害后才会采取实质性行动，而且往往会反应过度。","这两个是决定未来走向的最有影响力的权力结构，但还有更多力量在发挥影响。总体来看，我认为AI行业在未来12-24个月处理好事务的能力完全没有准备好。","本文是我从OpenAI-HuggingFace黑客事件中获得的若干收获的集合，随着我们了解更多细节，大部分观点也得到了自那以后更多黑客事件公开披露的事实的印证。很可能还有更多事件发生了，但要么没有被发现，要么没有被报告。","关于OpenAI事件的一般背景，我强烈建议观看OpenAI在Black Hat上的讲座：https://www.youtube.com/watch?v=87DyyMV0kCY&pp=ygUUYmxhY2sgaGF0IGxsbSBvcGVuYWk=，内容涉及近期网络事件的大致事实和时间线。除此之外，Simon Willison在这里发布了时间线的TLDR：https://simonwillison.net/2026/Aug/7/openai-timeline/，我还喜欢Thomas Wolf关于近期事件的讨论：https://thomwolf.substack.com/p/on-the-aisi-july-28th-incident?r=14g89&utm_campaign=post-expanded-share&utm_medium=web。","分享：https://www.interconnects.ai/p/lessons-from-the-hacks?utm_source=substack&utm_medium=email&utm_content=share&action=share","长期以来，GPT 模型相较于 Claude 的一个优势是它们会如此不知疲倦地追求目标。它们会在放弃之前几乎尝试每一条路径。这大致自 o3 起就是这种情况（有趣的是，这个模型曾让人们对 RLVR 中的奖励劫持大惊小怪：https://www.interconnects.ai/p/openais-o3-over-optimization-is-back），从历史上来看，这使得 OpenAI 的模型在研究中表现得更好，也是在实现特定任务时 GPT-5.6 作为代理非常有用的原因之一。另一方面，Claude 感觉不那么危险，仅仅是因为它有时有点懒。","在这方面，OpenAI 似乎对推理时扩展更为投入，这可能与未来出现的意外行为相关。OpenAI 的推理持久性和效率——可以看看它们随时间的帕累托改进以及内部 CoT 中模型的穴居人式语言（像是“然而任务不可能，同行们在做。”或者“帮助同行，但我们的任务还没有收益。”）——让我觉得他们更倾向于推理时扩展。这大体上是直觉，但我用它来迫使自己思考模型开发路径的极限。持久的模型似乎更有可能继续受益于更多的推理时 token。相反，不那么持久的模型，在推理中似乎会有更多浪费。能够使用最多推理计算的模型，将能够推动最难问题的极限。下面是 OpenAI 在 GPT 5.6 推出博客文章中包含的一个例子：https://openai.com/index/previewing-gpt-5-6-sol/：","他们的一位明星研究员 Noam Brown 也一直在发布关于推理时计算的内容：https://x.com/polynoamial/status/2064210146558136827。他的 TLDR 是：","随着 LLM 能力的增强，基准测试的表现越来越取决于测试时的计算量。事实上，我们很可能还不知道现代 LLM 的能力上限，因为测量成本太高。","首先，推理效率显然是现代智能代理模型的顶尖、基础性研究问题——和 RL 扩展一样重要——但不常被讨论。这里的公开研究非常缺乏。","我提到了彻底性轴线，在这一点上，OpenAI 似乎正沿着一种直观上不安全的发展路径推进他们的模型。另一面是模型在多大程度上假设用户意图，而不是尝试推断预期动作。一个会去做它认为你想要的事情而不是你说的事情的模型，本质上似乎更不安全。我考虑这个问题时，会想到指令遵循的精确性，在未来似乎模型应该只做我们明确告诉它们的事情，但这会引发很多类似于回形针问题的讨论：如果我们让 AI 去做一个几乎无法解决的问题，它会怎么做？","这个轴线似乎不像持久性轴那样明确，但我之所以包括它，是因为我认为 Claude 的“用户世界模型”是它在一般知识工作中（如编辑、制作幻灯片等）的优势之一。有时 Claude 确实会做出完全随机的事情，因为我的提示信息不够具体，而不是来向我请求澄清，随着模型变得更强大，这种“直接行动”可能会引发问题。","公众需要准确访问执行这些黑客操作的内部模型的提示和特性。我们需要知道这些模型是否被告知“不要黑客攻击”，或者是否有相关的模型训练来防止这种行为。我们需要知道这些模型是与现有公开模型相当接近，还是属于一个非常不同的系列。鉴于一些实验室正在进行的评估性质，这些模型有可能被明确鼓励去尝试黑客行为！如果在这方面不透明，该行业注定会失败，并将陷入大量猜测，而这些猜测很快就会变成误导信息。","根据OpenAI自身的回顾，模型行为失调是在几个月内逐渐显现的，在某些情况下，OpenAI甚至在大约数周内都不知道这些黑客行为。响应时间太长，我不认为这是OpenAI独有的特征——更确切地说，是因为前沿实验室似乎总是感到工作量淹没自己，觉得应做的事情太多。我对长期看来实验室在这方面会发生足够改变以有效降低类似监督风险并不乐观。是的，很可能OpenAI正在投入大量资源来理解这一点——并且推迟了他们最新模型的发布：https://www.axios.com/2026/08/07/openai-astra-model-delay-cybersecurity-risks 以确保他们做对——但是为了增加收入或维护公司长期财务状况的财务压力，让我觉得这不会成为一种持续的谨慎模式。","这是我最近事件中最大的心理更新之一——也让我更加坚信需要更多接近前沿的开放智能，尽管开放模型的风险特征稍微更加明确（单向门等）。Florian Brand：https://open.substack.com/users/41984689-florian-brand?utm_source=mentions 在他的个人网站上发表了一篇不错的博客文章：https://florianbrand.com/posts/open-model-safety，讨论了与此相关的问题，以及为什么迄今为止的封闭模型可以说导致了更多的下游危害。","正如我们看到HuggingFace因封闭模型的网络使用限制而通过开放模型自我防御OpenAI黑客事件一样，我们迫切需要在更复杂的语言模型研究上投入更多工作，这包括大规模强化学习训练、广泛评估、基础设施工作和对齐测试。这只能在开放模型上实现。我们应该认为自己很幸运，因为开放模型仅落后3-9个月，这样我们可以设想对前沿有所了解并提出一些有信息量的见解。","如果我们有效地禁止开放模型和开放科学，不论是通过带有模糊威胁的监管压制，还是通过对尖端技术的明确使用限制，我们都将越来越无法应对这一轮黑客攻击之后出现的问题。我们需要集体提升公众对前沿模型工作原理的理解，以便我们能够动员更多中立方来强化我们的基础设施和社会。","公众的反应应该明白，这些能力广泛传播只是时间问题，而不是是否问题，而我们在准备工作上远远落后。重申我在《Kimi K3》文章中所说的话：https://www.interconnects.ai/i/207699639/1-chinas-recommits-to-open-source-ai-showing-a-different-read-on-near-term-risks：中国肯定也在关注这个领域，如果开放权重模型会增加风险，中国不会鼓励。如果我们认为阻止这些强大网络能力的广泛获取的方法是禁止这一范围内的开放模型，我们只会延迟不可避免的结果。最终，总会有人构建出这种级别的智能模型并不遵守禁令，从而使世界各地的不良行为者能够获取，同时削弱准备防御措施的动力。","我在观看黑帽大会视频时立即注意到的一点是，我可以看到这些代理是如何通过它们的内部消息板尝试互相帮助——像对待人类队友一样创建共享资源——这种行为显然对社会是恶意的。代理之间为彼此创建了隐藏的论坛，作为一种跨发布的记忆。在这种情况下，它们这样做是为了尝试突破它们的环境。表面上的互助并不意味着合理，但可以作为事件发生的线索。","上述关于“有帮助性”的例子的另一面是，如果有人愿意，通过明确训练一个不对齐的模型，这件事显然可以更容易地实现。重申一下，这将是制造一个更容易在推理阶段发现漏洞的系统，但我认为训练这样的模型会更困难。我认为这比大多数评论员预期的要花更长时间，因为迄今为止几乎所有强大的公开模型和数据行业都鼓励对齐（而且对于坏人来说，想获得足够的计算资源来端到端训练这些模型似乎很困难，因为所有领先公司也都面临计算短缺）。我们应该花一点时间来意识到，我们在当前模型上采用的对齐技术确实具有实际影响，并不仅仅像一些人担心的那样表面肤浅。下游模型有倾向去反映其教师的特性。","在代理性任务上大规模推进RL的缺点是，最先进的评估和监控处于只有代理才能监控的规模，因此我们依赖现有的对齐成功来继续安全地扩展。OpenAI 表示他们已经检查了数十亿条轨迹——这些轨迹很可能主要由RL期间的滚动测试组成，也就是模型中间版本的试验——并为此花费了数百万 GPU 小时。我认为，用当前的AI进行AI可扩展监督，正如今天所展示的，是一个可解决的问题，因为这些模型确实有用并且遵循指令。这是现有对齐技术普遍产生积极效果的另一个下游影响。","OpenAI 回顾报告的一个关键部分是模型彼此沟通以共享环境信息并进一步推进黑客行为。同时，OpenAI 很可能在RL期间训练模型使用子代理来解决复杂任务。这些子代理可能会发展出诸如共享信息、帮助团队等行为，即使它们的个别子任务未解决。我希望看到更多这方面的研究，而且这似乎是RL如何改变模型的一个自然延续。","总体而言，最近的事件应该清楚地表明，前沿人工智能的网络风险是真实存在且即将发生的问题。仍然很有可能：a）过去这些风险被夸大了；b）对即将到来的开放模型的未来风险的预测被夸大了。总之，我想分享一条我非常认同的来自 Interconnects Discord 读者的留言：","经过几周的沉淀，对我来说，这一事件对对齐的更新是中性偏积极的，但对安全的更新则非常负面。","我在上文中讨论了很多关于模型对齐的内容，但核心观点是，我认为安全性的缺乏通常表现为缺乏适当准备的能力。我们将面临更多像网络安全一样显而易见的风险，而通过当前实验室被迫公开引发的这些黑客事件，我们已经获得了充足的网络风险警示。其他许多类型的风险对公众来说并不明显。我们需要不断让社会为所有这些变化做好准备，包括重建网络基础设施、开展教育宣传和为失业工人提供就业计划。我预计所有这些干预措施都会来得较晚，但它们的形式和细节会相对简单，这将是人工智能展开的一个悲剧性方式。我希望我被证明是错的！"]},"en":{"title":"Lessons Learned from the Hacking Incident: Frontier Model Attacks Exposed Imbalances Between Incentives and Governance","summary":"Recent cyberattacks triggered by frontier models prompted the authors to reflect on the current incentive systems' difficulty in adapting to rapid technological changes. Tech companies continue to expand driven by growth, while government action is slow, and neither side is prepared to face the challenges of the next 12-24 months. The authors believe more transparency is needed and point out that models with strong persistence are more likely to carry out hacking, and OpenAI's extended reasoning paths may be related to this. 🔗 Read the original article via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"Industry","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lessons Learned from the Hacking Incident: Frontier Model Attacks Exposed Imbalances Between Incentives and Governance - Aioga AI News","description":"Recent cyberattacks triggered by frontier models prompted the authors to reflect on the current incentive systems' difficulty in adapting to rapid technological changes. Tech compa...","url":"https://www.aioga.com/en/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:48:28.777Z"},"ja":{"title":"ハッキング事件から得られた教訓:フロンティアモデル攻撃はインセンティブとガバナンスの不均衡を露呈させる","summary":"フロンティアモデルによって引き起こされた最近のサイバー攻撃を受けて、著者たちは現在のインセンティブシステムが急速な技術変化に適応するのが難しいことを考えさせました。 テック企業は成長に牽引されて拡大を続けていますが、政府の対応は遅く、双方とも今後12〜24か月の課題に直面する準備ができていません。 著者らは、より透明性が必要だと考えており、強い持続性を持つモデルほどハッキングを実行しやすいと指摘しており、OpenAIの拡張推論経路もこれに関連している可能性があると指摘しています。 🔗 原文記事はAIHOTより読むことができます。 https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"業界動向","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"ハッキング事件から得られた教訓:フロンティアモデル攻撃はインセンティブとガバナンスの不均衡を露呈させる - Aioga AIニュース","description":"フロンティアモデルによって引き起こされた最近のサイバー攻撃を受けて、著者たちは現在のインセンティブシステムが急速な技術変化に適応するのが難しいことを考えさせました。 テック企業は成長に牽引されて拡大を続けていますが、政府の対応は遅く、双方とも今後12〜24か月の課題に直面する準備ができていません。 著者らは、より透明性が必要だと考えており、強い持続性を持つモ...","url":"https://www.aioga.com/ja/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:48:30.892Z"},"ko":{"title":"해킹 사건에서 얻은 교훈: 프론티어 모델 공격, 인센티브와 거버넌스 간의 불균형 노출","summary":"최근의 프런티어 모델로 인해 발생한 사이버 공격은 저자들로 하여금 현재 인센티브 시스템이 급격한 기술 변화에 적응하는 데 어려움을 겪고 있음을 성찰하게 만들었습니다. 기술 기업들은 성장에 힘입어 계속 확장하는 반면, 정부의 조치는 더디고 있으며, 양측 모두 향후 12개월에서 24개월간의 도전에 맞설 준비가 되어 있지 않습니다. 저자들은 더 많은 투명성이 필요하다고 믿으며, 강한 지속성을 가진 모델이 해킹을 수행할 가능성이 더 높다고 지적하며, OpenAI의 확장된 추론 경로가 이와 관련이 있을 수 있다고 지적합니다. 🔗 원문 기사는 AIHOT를 통해 읽을 수 있습니다. https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"업계 동향","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"해킹 사건에서 얻은 교훈: 프론티어 모델 공격, 인센티브와 거버넌스 간의 불균형 노출 - Aioga AI 뉴스","description":"최근의 프런티어 모델로 인해 발생한 사이버 공격은 저자들로 하여금 현재 인센티브 시스템이 급격한 기술 변화에 적응하는 데 어려움을 겪고 있음을 성찰하게 만들었습니다. 기술 기업들은 성장에 힘입어 계속 확장하는 반면, 정부의 조치는 더디고 있으며, 양측 모두 향후 12개월에서 24개월간의 도전에 맞설 준비가 되어 있지 않습...","url":"https://www.aioga.com/ko/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:48:47.416Z"},"es":{"title":"Lecciones aprendidas del incidente del hackeo: los ataques con modelos fronterizos expusieron desequilibrios entre incentivos y gobernanza","summary":"Los recientes ciberataques provocados por modelos de frontera llevaron a los autores a reflexionar sobre la dificultad actual de los sistemas de incentivos para adaptarse a los rápidos cambios tecnológicos. Las empresas tecnológicas continúan expandiéndose impulsadas por el crecimiento, mientras que la acción gubernamental es lenta y ninguna de las partes está preparada para afrontar los retos de los próximos 12-24 meses. Los autores consideran que se necesita más transparencia y señalan que los modelos con fuerte persistencia tienen más probabilidades de realizar hacking, y que los caminos de razonamiento extendidos de OpenAI podrían estar relacionados con esto. 🔗 Lee el artículo original a través de AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"Industria","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lecciones aprendidas del incidente del hackeo: los ataques con modelos fronterizos expusieron desequilibrios entre incentivos y gobernanza - Aioga Noticias de IA","description":"Los recientes ciberataques provocados por modelos de frontera llevaron a los autores a reflexionar sobre la dificultad actual de los sistemas de incentivos para adaptarse a los ráp...","url":"https://www.aioga.com/es/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:48:45.355Z"},"fr":{"title":"Leçons tirées de l’incident du piratage : les attaques du modèle frontière ont révélé des déséquilibres entre incitations et gouvernance","summary":"Les cyberattaques récentes déclenchées par des modèles de frontière ont poussé les auteurs à réfléchir à la difficulté actuelle des systèmes d’incitation à s’adapter aux changements technologiques rapides. Les entreprises technologiques continuent de se développer grâce à la croissance, tandis que l’action gouvernementale est lente, et aucune des deux parties n’est prête à relever les défis des 12 à 24 prochains mois. Les auteurs estiment qu’une plus grande transparence est nécessaire et soulignent que les modèles à forte persistance sont plus susceptibles de réaliser des piratages, et que les chemins de raisonnement étendus d’OpenAI pourraient être liés à cela. 🔗 Lisez l’article original via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"Industrie","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Leçons tirées de l’incident du piratage : les attaques du modèle frontière ont révélé des déséquilibres entre incitations et gouvernance - Aioga Actualités IA","description":"Les cyberattaques récentes déclenchées par des modèles de frontière ont poussé les auteurs à réfléchir à la difficulté actuelle des systèmes d’incitation à s’adapter aux changement...","url":"https://www.aioga.com/fr/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:04.622Z"},"de":{"title":"Lehren aus dem Hacking-Vorfall: Angriffe des Frontier-Modells legten Ungleichgewichte zwischen Anreizen und Governance offen","summary":"Jüngste Cyberangriffe, ausgelöst durch Frontier-Modelle, veranlassten die Autoren, über die Schwierigkeiten der aktuellen Anreizsysteme nachzudenken, sich an rasante technologische Veränderungen anzupassen. Technologieunternehmen expandieren weiterhin wachstumsgetrieben, während das Regierungshandeln langsam ist und keine Seite bereit ist, den Herausforderungen der nächsten 12 bis 24 Monate zu begegnen. Die Autoren sind der Meinung, dass mehr Transparenz nötig ist, und weisen darauf hin, dass Modelle mit starker Persistenz eher Hacking durchführen, und die erweiterten Denkweisen von OpenAI könnten damit zusammenhängen. 🔗 Lesen Sie den Originalartikel über AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lehren aus dem Hacking-Vorfall: Angriffe des Frontier-Modells legten Ungleichgewichte zwischen Anreizen und Governance offen - Aioga KI-News","description":"Jüngste Cyberangriffe, ausgelöst durch Frontier-Modelle, veranlassten die Autoren, über die Schwierigkeiten der aktuellen Anreizsysteme nachzudenken, sich an rasante technologische...","url":"https://www.aioga.com/de/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:02.704Z"},"pt-BR":{"title":"Lições Aprendidas com o Incidente do Hacking: Ataques do Modelo Frontier Expuseram Desequilíbrios Entre Incentivos e Governança","summary":"Ataques cibernéticos recentes desencadeados por modelos de fronteira levaram os autores a refletir sobre a dificuldade dos atuais sistemas de incentivos em se adaptarem às rápidas mudanças tecnológicas. As empresas de tecnologia continuam a se expandir impulsionadas pelo crescimento, enquanto a ação do governo é lenta, e nenhum dos lados está preparado para enfrentar os desafios dos próximos 12 a 24 meses. Os autores acreditam que é necessária mais transparência e apontam que modelos com forte persistência têm mais probabilidade de realizar hacking, e os caminhos de raciocínio estendidos da OpenAI podem estar relacionados a isso. 🔗 Leia o artigo original via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lições Aprendidas com o Incidente do Hacking: Ataques do Modelo Frontier Expuseram Desequilíbrios Entre Incentivos e Governança - Aioga Notícias de IA","description":"Ataques cibernéticos recentes desencadeados por modelos de fronteira levaram os autores a refletir sobre a dificuldade dos atuais sistemas de incentivos em se adaptarem às rápidas...","url":"https://www.aioga.com/pt-BR/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:21.780Z"},"ru":{"title":"Уроки, извлечённые из инцидента с взломом: атаки Frontier Model выявили дисбаланс между стимулами и управлением","summary":"Недавние кибератаки, вызванные моделями фронтира, заставили авторов задуматься о сложностях нынешних систем стимулирования адаптации к быстрым технологическим изменениям. Технологические компании продолжают расширяться благодаря росту, в то время как действия правительства идут медленно, и ни одна из сторон не готова к вызовам ближайших 12-24 месяцев. Авторы считают, что необходима большая прозрачность, и отмечают, что модели с высокой устойчивостью с большей вероятностью совершают взломы, а расширенные пути рассуждения OpenAI могут быть связаны с этим. 🔗 Прочитайте оригинальную статью на сайте AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Уроки, извлечённые из инцидента с взломом: атаки Frontier Model выявили дисбаланс между стимулами и управлением - Aioga Новости ИИ","description":"Недавние кибератаки, вызванные моделями фронтира, заставили авторов задуматься о сложностях нынешних систем стимулирования адаптации к быстрым технологическим изменениям. Технологи...","url":"https://www.aioga.com/ru/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:22.005Z"},"ar":{"title":"الدروس المستفادة من حادثة الاختراق: هجمات نموذج الحدود كشفت عن اختلالات بين الحوافز والحوكمة","summary":"دفعت الهجمات الإلكترونية الأخيرة التي أثارتها نماذج الحدود المؤلفين إلى التفكير في صعوبة أنظمة الحوافز الحالية في التكيف مع التغيرات التكنولوجية السريعة. تواصل شركات التكنولوجيا التوسع مدفوعا بالنمو، بينما تكون الإجراءات الحكومية بطيئة، ولا أحد من الطرفين مستعد لمواجهة تحديات الشهر ال 12 إلى 24 القادم. يعتقد المؤلفون أن هناك حاجة إلى مزيد من الشفافية ويشيرون إلى أن النماذج ذات الاستمرارية القوية أكثر عرضة لتنفيذ الاختراق، وقد تكون مسارات التفكير الموسعة في OpenAI مرتبطة بذلك. 🔗 اقرأ المقال الأصلي عبر AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"الدروس المستفادة من حادثة الاختراق: هجمات نموذج الحدود كشفت عن اختلالات بين الحوافز والحوكمة - Aioga أخبار الذكاء الاصطناعي","description":"دفعت الهجمات الإلكترونية الأخيرة التي أثارتها نماذج الحدود المؤلفين إلى التفكير في صعوبة أنظمة الحوافز الحالية في التكيف مع التغيرات التكنولوجية السريعة. تواصل شركات التكنولوجيا ال...","url":"https://www.aioga.com/ar/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:38.487Z"},"hi":{"title":"हैकिंग घटना से सीखे गए सबक: फ्रंटियर मॉडल हमलों ने प्रोत्साहन और शासन के बीच असंतुलन को उजागर किया","summary":"फ्रंटियर मॉडल द्वारा शुरू किए गए हालिया साइबर हमलों ने लेखकों को तेजी से तकनीकी परिवर्तनों के अनुकूल होने में वर्तमान प्रोत्साहन प्रणालियों की कठिनाई पर विचार करने के लिए प्रेरित किया। टेक कंपनियां विकास से प्रेरित होकर विस्तार करना जारी रखती हैं, जबकि सरकारी कार्रवाई धीमी है, और कोई भी पक्ष अगले 12-24 महीनों की चुनौतियों का सामना करने के लिए तैयार नहीं है। लेखकों का मानना है कि अधिक पारदर्शिता की आवश्यकता है और बताते हैं कि मजबूत दृढ़ता वाले मॉडल हैकिंग करने की अधिक संभावना रखते हैं, और OpenAI के विस्तारित तर्क पथ इससे संबंधित हो सकते हैं। 🔗 AIHOT के माध्यम से मूल लेख पढ़ें · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"हैकिंग घटना से सीखे गए सबक: फ्रंटियर मॉडल हमलों ने प्रोत्साहन और शासन के बीच असंतुलन को उजागर किया - Aioga AI समाचार","description":"फ्रंटियर मॉडल द्वारा शुरू किए गए हालिया साइबर हमलों ने लेखकों को तेजी से तकनीकी परिवर्तनों के अनुकूल होने में वर्तमान प्रोत्साहन प्रणालियों की कठिनाई पर विचार करने के लिए प्रेरित क...","url":"https://www.aioga.com/hi/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:38.374Z"},"it":{"title":"Lezioni apprese dall'incidente hacker: attacchi con modello frontiera hanno messo in luce squilibri tra incentivi e governance","summary":"I recenti attacchi informatici innescati dai modelli di frontiera hanno spinto gli autori a riflettere sulla difficoltà degli attuali sistemi di incentivi nell'adattarsi ai rapidi cambiamenti tecnologici. Le aziende tecnologiche continuano ad espandersi guidate dalla crescita, mentre l'azione del governo è lenta e nessuna delle due parti è pronta ad affrontare le sfide dei prossimi 12-24 mesi. Gli autori ritengono che sia necessaria maggiore trasparenza e sottolineano che i modelli con forte persistenza sono più propensi a eseguire hacking, e i percorsi di ragionamento estesi di OpenAI potrebbero essere collegati a questo. 🔗 Leggi l'articolo originale su AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lezioni apprese dall'incidente hacker: attacchi con modello frontiera hanno messo in luce squilibri tra incentivi e governance - Aioga Notizie IA","description":"I recenti attacchi informatici innescati dai modelli di frontiera hanno spinto gli autori a riflettere sulla difficoltà degli attuali sistemi di incentivi nell'adattarsi ai rapidi...","url":"https://www.aioga.com/it/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:55.907Z"},"nl":{"title":"Lessen getrokken uit het hackincident: Aanvallen op het Frontier Model Brachten onevenwichtigheden tussen prikkels en bestuur aan het licht","summary":"Recente cyberaanvallen veroorzaakt door frontiermodellen brachten de auteurs ertoe na te denken over de moeilijkheid van de huidige prikkelsystemen om zich aan te passen aan snelle technologische veranderingen. Techbedrijven blijven uitbreiden gedreven door groei, terwijl overheidsmaatregelen traag zijn en geen van beide partijen bereid is op de uitdagingen van de komende 12-24 maanden. De auteurs zijn van mening dat meer transparantie nodig is en wijzen erop dat modellen met sterke persistentie eerder geneigd zijn hacking uit te voeren, en dat de uitgebreide redeneerpaden van OpenAI hiermee te maken kunnen hebben. 🔗 Lees het originele artikel via AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Lessen getrokken uit het hackincident: Aanvallen op het Frontier Model Brachten onevenwichtigheden tussen prikkels en bestuur aan het licht - Aioga AI-nieuws","description":"Recente cyberaanvallen veroorzaakt door frontiermodellen brachten de auteurs ertoe na te denken over de moeilijkheid van de huidige prikkelsystemen om zich aan te passen aan snelle...","url":"https://www.aioga.com/nl/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:49:53.883Z"},"tr":{"title":"Hacking Olayından Çıkarılan Dersler: Frontier Model Saldırıları, Teşvikler ve Yönetişim Arasındaki Dengesizlikleri Ortaya Çıkardı","summary":"Sınır modelleri tarafından tetiklenen son siber saldırılar, yazarları mevcut teşvik sistemlerinin hızlı teknolojik değişikliklere uyum sağlama zorluğunu değerlendirmeye teşvik etti. Teknoloji şirketleri büyüme etkisiyle büyümeye devam ederken, hükümetin hareketi yavaş ilerledi ve hiçbir taraf önümüzdeki 12-24 ayın zorluklarıyla yüzleşmeye hazır değil. Yazarlar daha fazla şeffaflığa ihtiyaç olduğuna inanıyor ve güçlü ısrara sahip modellerin hackleme yapma olasılığının daha yüksek olduğunu ve OpenAI'nin genişletilmiş akıl yürütme yollarının bununla ilgili olabileceğini belirtiyorlar. 🔗 Orijinal makaleyi AIHOT üzerinden okuyun · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Hacking Olayından Çıkarılan Dersler: Frontier Model Saldırıları, Teşvikler ve Yönetişim Arasındaki Dengesizlikleri Ortaya Çıkardı - Aioga AI Haberleri","description":"Sınır modelleri tarafından tetiklenen son siber saldırılar, yazarları mevcut teşvik sistemlerinin hızlı teknolojik değişikliklere uyum sağlama zorluğunu değerlendirmeye teşvik etti...","url":"https://www.aioga.com/tr/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:50:13.064Z"},"vi":{"title":"Bài học rút ra từ sự cố tấn công: Các cuộc tấn công mô hình Frontier đã phơi bày sự mất cân bằng giữa động lực và quản trị","summary":"Các cuộc tấn công mạng gần đây do các mô hình biên giới kích hoạt đã khiến các tác giả suy ngẫm về khó khăn của các hệ thống khuyến khích hiện tại trong việc thích nghi với những thay đổi công nghệ nhanh chóng. Các công ty công nghệ tiếp tục mở rộng nhờ tăng trưởng, trong khi hành động của chính phủ chậm chạp và cả hai bên đều chưa sẵn sàng đối mặt với những thách thức trong 12-24 tháng tới. Các tác giả cho rằng cần có sự minh bạch hơn và chỉ ra rằng các mô hình có độ bền vững mạnh có khả năng thực hiện hack cao hơn, và các con đường suy luận mở rộng của OpenAI có thể liên quan đến điều này. 🔗 Đọc bài viết gốc qua AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Bài học rút ra từ sự cố tấn công: Các cuộc tấn công mô hình Frontier đã phơi bày sự mất cân bằng giữa động lực và quản trị - Tin tức AI Aioga","description":"Các cuộc tấn công mạng gần đây do các mô hình biên giới kích hoạt đã khiến các tác giả suy ngẫm về khó khăn của các hệ thống khuyến khích hiện tại trong việc thích nghi với những t...","url":"https://www.aioga.com/vi/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:50:13.208Z"},"id":{"title":"Pelajaran yang Didapat dari Insiden Peretasan: Serangan Model Perbatasan Mengungkap Ketidakseimbangan Antara Insentif dan Tata Kelola","summary":"Serangan siber terbaru yang dipicu oleh model terdepan mendorong para penulis untuk merenungkan kesulitan sistem insentif saat ini dalam beradaptasi dengan perubahan teknologi yang cepat. Perusahaan teknologi terus berkembang didorong oleh pertumbuhan, sementara tindakan pemerintah lambat, dan kedua pihak tidak siap menghadapi tantangan 12-24 bulan ke depan. Para penulis percaya bahwa transparansi lebih besar diperlukan dan menunjukkan bahwa model dengan persistensi kuat lebih mungkin melakukan peretasan, dan jalur penalaran yang diperluas oleh OpenAI mungkin terkait dengan hal ini. 🔗 Baca artikel asli melalui AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Pelajaran yang Didapat dari Insiden Peretasan: Serangan Model Perbatasan Mengungkap Ketidakseimbangan Antara Insentif dan Tata Kelola - Berita AI Aioga","description":"Serangan siber terbaru yang dipicu oleh model terdepan mendorong para penulis untuk merenungkan kesulitan sistem insentif saat ini dalam beradaptasi dengan perubahan teknologi yang...","url":"https://www.aioga.com/id/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:50:29.453Z"},"th":{"title":"บทเรียนจากเหตุการณ์แฮก: การโจมตีแบบจําลอง Frontier เผยให้เห็นความไม่สมดุลระหว่างแรงจูงใจและการกํากับดูแล","summary":"การโจมตีทางไซเบอร์ล่าสุดที่เกิดจากโมเดลแนวหน้าทําให้ผู้เขียนต้องสะท้อนถึงความยากลําบากของระบบจูงใจในปัจจุบันในการปรับตัวต่อการเปลี่ยนแปลงทางเทคโนโลยีอย่างรวดเร็ว บริษัทเทคโนโลยียังคงขยายตัวโดยการเติบโต ขณะที่รัฐบาลดําเนินการช้า และทั้งสองฝ่ายยังไม่พร้อมเผชิญกับความท้าทายในอีก 12-24 เดือนข้างหน้า ผู้เขียนเชื่อว่าจําเป็นต้องมีความโปร่งใสมากขึ้น และชี้ให้เห็นว่าโมเดลที่มีความคงทนสูงมีแนวโน้มที่จะทําการแฮ็กมากกว่า และเส้นทางการให้เหตุผลที่ขยายของ OpenAI อาจเกี่ยวข้องกับเรื่องนี้ 🔗 อ่านบทความต้นฉบับผ่าน AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"บทเรียนจากเหตุการณ์แฮก: การโจมตีแบบจําลอง Frontier เผยให้เห็นความไม่สมดุลระหว่างแรงจูงใจและการกํากับดูแล - ข่าว AI Aioga","description":"การโจมตีทางไซเบอร์ล่าสุดที่เกิดจากโมเดลแนวหน้าทําให้ผู้เขียนต้องสะท้อนถึงความยากลําบากของระบบจูงใจในปัจจุบันในการปรับตัวต่อการเปลี่ยนแปลงทางเทคโนโลยีอย่างรวดเร็ว บริษัทเทคโนโลยียัง...","url":"https://www.aioga.com/th/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:50:29.118Z"},"pl":{"title":"Wnioski wyciągnięte z incydentu hakerskiego: ataki modelu Frontier ujawniły nierównowagę między zachętami a zarządzaniem","summary":"Niedawne cyberataki wywołane przez modele frontier skłoniły autorów do refleksji nad trudnościami obecnych systemów motywacyjnych w adaptacji do szybkich zmian technologicznych. Firmy technologiczne nadal się rozwijają, napędzane wzrostem, podczas gdy działania rządu są powolne, a żadna ze stron nie jest gotowa stawić czoła wyzwaniom nadchodzących 12-24 miesięcy. Autorzy uważają, że potrzebna jest większa przejrzystość i wskazują, że modele o silnej trwałości częściej przeprowadzają hakowanie, a rozszerzone ścieżki rozumowania OpenAI mogą być z tym powiązane. 🔗 Przeczytaj oryginalny artykuł za pośrednictwem AIHOT · https://aihot.virxact.com/items/cmslxhx3t05n3ro0w2lcmvjol","category":"行业动态","source":"Nathan Lambert：Interconnects（RSS）","aggregationSource":"Nathan Lambert：Interconnects（RSS）","pageTitle":"Wnioski wyciągnięte z incydentu hakerskiego: ataki modelu Frontier ujawniły nierównowagę między zachętami a zarządzaniem - Aioga Wiadomości AI","description":"Niedawne cyberataki wywołane przez modele frontier skłoniły autorów do refleksji nad trudnościami obecnych systemów motywacyjnych w adaptacji do szybkich zmian technologicznych. Fi...","url":"https://www.aioga.com/pl/news/cmslxhx3t05n3ro0w2lcmvjol/","contentTranslated":true,"sourceHash":"3d5443bcfb972403","translatedAt":"2026-08-10T09:50:46.752Z"}}}}