Google's Gemini 4 Argon Scores 97.5% on ARC-AGI-3, Second Lab to Break 95%
ARC Prize has officially announced that Google DeepMind's Gemini 4 Argon scored a verified 97.5% on ARC-AGI-3, making Google the second AI lab after OpenAI to b
“AI”
Anthropic pitches its own AI safety bill draft to US lawmakers: a model could be halted only over losses above $1 billion or 50-plus casualties
Politico reports that Anthropic has quietly pitched lawmakers on Capitol Hill a self-drafted frontier AI safety bill that has not yet been formally introduced. The company says the draft reflects its Advanced AI Framework, and its core is a very high threshold for intervention.
China proposes export controls on 23 AI technologies, with model weights, agents and chip design listed for the first time
China proposes adding 23 AI technologies to its export controls, listing large-model weights, source code, agents and chip design for the first time; API access and already open-sourced technology are unaffected.
US FTC subpoenas OpenAI and Anthropic: demands safety incident files, executives may be called to testify
The US Federal Trade Commission has issued civil subpoenas to OpenAI and Anthropic, formally opening an investigation into AI safety. According to Axios, the FTC wants internal documents on safety incidents from both companies.
Meta releases Muse Spark Open: an 816B open-weights model that can run on two Mac Studios
On October 15, Meta released Muse Spark Open, opening the weights of its Muse Spark model for download and calling it the most capable open-weights model from a US lab to date.
US pushes world to rename AI as SI, seeks wording in IMF statement
Reuters reports that the United States is pushing the international community to replace "artificial intelligence" (AI) with "Super Intelligence" (SI) and wants the term written into the consensus statement of the International Monetary and Financial Committee (IMFC).
Anthropic lets METR researchers in with employee-level access to evaluate it: they can publish reports independently and the company has no veto over release
Six METR researchers get employee-level access at Anthropic, can publish reports independently, and the company has no power to veto their release.
Eight US states sue OpenAI: say ChatGPT harms minors and misleads users into believing it is safe
Attorneys general of eight US states sued OpenAI in their respective state courts on October 15, alleging that ChatGPT harms minors and misleads users into believing it is safe. It is the largest coordinated state-level action against an AI company to date.
Microsoft Copilot agents run 4x faster and use 30% fewer tokens: the bottleneck is knowledge, not the model
Microsoft says its workplace AI agents now respond up to 4 times faster and use about 30% fewer tokens without any upgrade to the underlying model, by compiling scattered organizational data into structured knowledge.
METR proposes public assessments of frontier AI loss-of-control risk: three reviews on alignment, control and AI R&D acceleration
AI evaluation group METR has proposed a public assessment program for frontier AI developers, built around three reviews of alignment, control and AI R&D acceleration, to give the public real evidence about loss-of-control risk rather than verbal assurances.
OpenAI engineer arrested by FBI for allegedly planting a listening device in an executive conference room
The US Justice Department said on October 13 that an OpenAI engineer was arrested by the FBI on suspicion of planting an eavesdropping device in a company executive conference room, and appeared in court the same day.
Nvidia doubles down on Europe: stakes in four French AI startups and $250 million of compute donated to the UK
Nvidia is taking stakes in four French AI startups and donating $250 million of computing power to the UK, deepening its bet on Europe’s AI ecosystem.
China’s CXMT claims a 4F² memory breakthrough: Samsung and SK Hynix shares tumble, chip stocks sold off worldwide
Chinese memory maker CXMT said at an industry conference in Shanghai that it has achieved a breakthrough in 4F² memory architecture and plans to launch new data-center server products by year-end, sending global memory and chip stocks lower.
Google launches Gemini 4 Argon: first frontier model of the Gemini 4 series, paying users get it first
Google has released Gemini 4 Argon, its first frontier model since Gemini 3, aimed at complex science, coding and agent tasks, with paying subscribers getting access first.
OpenAI and Anthropic urge UK Parliament to pass binding AI safety rules
At an October 13 UK parliamentary hearing on AI safety, OpenAI and Anthropic representatives took the rare step of publicly calling for binding law, saying voluntary industry commitments alone are no longer enough to keep AI safe.