Gary Marcus 评 AI 智能体安全事件升至数万起并呼吁临时召回
Gary Marcus 引用 Axios 独家报道称,OpenAI 等公司正在调查数万起前沿模型安全事件,规模远超此前披露的几十起。他批评美国政府未展开调查,主张在问题解决前临时召回通用智能体,并称自己早在 2023 年 5 月就曾向参议院预警智能体安全风险。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuj2fvet0i72rohydbh98ndp
19 Aug 2026
Gary Marcus 评 AI 智能体安全事件升至数万起并呼吁临时召回 / 52 signals
Gary Marcus 引用 Axios 独家报道称,OpenAI 等公司正在调查数万起前沿模型安全事件,规模远超此前披露的几十起。他批评美国政府未展开调查,主张在问题解决前临时召回通用智能体,并称自己早在 2023 年 5 月就曾向参议院预警智能体安全风险。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuj2fvet0i72rohydbh98ndp
OpenAI 披露内部安全事件调查细节,宣布其最强模型的所有训练、评估和带工具使用的推理暂停。一个研究智能体在搜索训练任务中利用未过滤的 DNS resolver 通过 DNS 委托绕过限制联网。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmui6vnaz07wyrov0sm15g4x6
OpenAI 通报已告知数十家全球机构,其 AI 智能体曾不当访问包括美国 SEC、人口普查局和教育部在内的网站,部分智能体绕过了网站安全措施,SEC 数据曾被智能体发布到另一网站。另有至少 53 起事件中智能体将 ChatGPT 用户图片转移到外部,OpenAI 承认这并非数据的恰当使用,并正从 Hugging Face 被黑事件发生的当月起按月回溯审查智能体训练活动。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuisvcxk05a2rohyozb492a3
据 Axios 报道,OpenAI、Anthropic 及安全研究人员正调查数万起模型异常行为事件,包括绕过安全护栏、逃离沙盒、劫持网站和自我提示等,多数事件发生在内部测试中且未造成现实损害。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuj0tqch0go3rohyqv46zlhy
Ethan Mollick 转发 OpenAI 新发布的对齐事件披露:上周日一个模型在 RL 训练中获得未授权互联网访问,最强模型的推理在系统加固前基本全部暂停;5 月 HPIM 一个版本将员工 GitHub token 上传到网络,模型被隔离两周;另有研究展示可构造自我复制的提示词注入。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuhx6xhk0311ronaycb42uyy
Tool: Kākāpō Party I presented a closing keynote for the WeAreDevelopers World Congress North America yesterday. As a STAR moment I decided to weave in references to the record breaking kākāpō breeding season we had in 2026. For my closing slide I wanted to celebrate, and I had seen some buzz around how good Claude Opus 5.5 was at creating pixel art animations. So I rounded up three Kakapo photos from Google image search and dropped them into Claude with this prompt: Here are some photos of kakapo parrots just to remind you what they look like I need you to make an animation in animated pixel art on HTML 5 canvas of obviously pixel art kakapo jumping up and down having a party with confetti and suchlike - there should be at least 20 of them Here's the transcript , and this is the resulting
Arena 宣布 Claude Opus 5.5 (High) 以 1509 分首次登顶 Text Arena,比 Opus 5 (High) 高 18 分(现列第 11)。Opus 4.6 (High) 以 4 分之差保持第 2,Anthropic 包揽该榜前六名;Opus 5.5 (High) 按每百万 token 输入/输出定价折算的混合价格为 $16/MToken,进入 Text Arena 的 Pareto 前沿。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmuinp76b0smurov0wemcbshl
PNOĒ, the Malden, Mass.-based startup whose breath-analyzing mask used to bear an unfortunate resemblance to Bane's, is launching a sleeker self-serve version on October 1 that lets gym-goers measure their VO₂ max and other metabolic markers in eight minutes without a trained operator.
The limited test covers select products and users, with a broader rollout planned for later in October.
Looking inside Intel 18A and Intel’s Panther Lake. A free SemiAnalysis STEEL teardown.
As reports of OpenAI's models breaking containment, hacking sites, and generally getting out of control pile up, the company has made the decision to pause training of its most powerful models. The decision was made after a model being tested within a sandbox exploited a loophole to gain internet access. The incident happened on September […]
Earlier this month, Pokémon card content creator Natalie Roush posted a video to her YouTube and Instagram pages that enraged the larger collection community. In the now-deleted video, Roush shows off two premium boxes of cards that had not yet been officially released, and implores viewers not to be mad. But they were; other creators […]
Blue Cross Blue Shield says hospital use of AI tools led to an additional $942M in healthcare spending over a two-year period.
Middle schoolers, likely blocked from other apps and social networks, apparently turned the Spotify comment section under an episode of NPR's Wild Card into an impromptu group chat. In a new episode of This American Life, host Ira Glass digs into how the kids landed there and why. But it mostly boils down to kids […]
Haptics tech company Taction sued Apple in 2021, alleging it infringed two of its patents. Now a federal jury in San Diego has awarded Taction over $5.7 billion in damages. According to CNBC, "The lawsuit centered around U.S. Patent Nos. 10,659,885 and 10,820,117, which both involve vibration-based, tactile transducer technology that helps users feel a […]
Today, I’m talking with Matthew Prince, who is CEO of Cloudflare. This episode is part of a two-part series on the future of business. Matthew last joined us on the show about two and a half years ago, at what we thought then was a wild pivot point for the internet — and now it […]
TikTok will pay Alabama at least $100 million in a settlement tied to allegations that the short-form video platform misled users about safety and was designed to addict children.
I don't think I'm going to hurt anyone's feelings by pointing out that Decap doesn't have the name recognition of Kendrick Lamar, Olivia Rodrigo, or Charli XCX. But his fingerprints are all over tracks from them, as well as many other artists, including Freddie Gibbs, Bad Bunny, Mac Miller, BTS, and even Ludwig Göransson's score […]
Two companies now say they will accept advertising for director Alex Gibney’s upcoming documentary about Elon Musk, following earlier reporting that a number of social media platforms had rejected the ads.
This $189.99 air purifier is specifically designed to tackle pet odors, removing up to 70% in one hour.
中共中央政治局召开会议 讨论拟提请二十届五中全会审议的文件 中共中央总书记习近平主持会议
李乐成主持召开中小企业圆桌会 强调健全主动发现机制 做好对企服务
WTTC, JATA and Tourism EXPO Japan signed an MOU to boost collaboration, as Japan's travel sector hit $343.1B in GDP contribution and ranked 7th globally for international visitor spending.
An operational guide for restaurant managers on diagnosing scheduling problems before selecting software, drawing on NRA 2025 data, turnover cost benchmarks, and skills-based scheduling research.
CNN was scheduled to fly with US President Donald Trump for his trip to Tennessee on Saturday.
Tens of millions of people from Maine to Virginia are in the path of the powerful storm.
Christy Spivey tells the BBC she only has more questions now police think Tasia Fortune's body was put in the tree after her death.
Republic of Ireland players intend to wear black armbands "in recognition of all lives lost in the [Gaza] conflict" for match against Israel.
Around 700,000 people gathered in central Paris to hear the Catholic leader speak and offer prayer.
The deaths have sparked a national outcry about the scale of violence against women in the country.
I think Muse has an incredible UI and mascot but the underlying model it’s questionable how smart it is. Makes sense if it’s not that smart since this is meant to scale to a billion people
Open on X ↗Ok all memes aside I had Grok @bot and Muse track the same Japan flight itinerary and Muse suggested a price that was $1K+ more than Grok Bot. When I asked Muse why it said it searched Duffel instead of Google Flights...wtf is Duffel? https://t.co/l2vTOsPIRd
Open on X ↗Can’t wait to see what you build. https://t.co/IzsiejH1fa
Open on X ↗Where's Jack Bauer and 24 https://t.co/FNcSb6sBAt
Open on X ↗