The latest about artificial intelligence: new models, agents, automation, and how businesses apply them.
ICLR 2027 has collected about 50,000 abstracts a week before the deadline, against roughly 19,500 valid submissions for ICLR 2026.
A July presentation puts OpenAI's compute and infrastructure spending at about $856bn by 2030, up from the roughly $600bn target shown to investors in February.
Altman, Huang and Cook are expected at the 24 September state dinner, five months after a Beijing meeting that produced no agreement and no chip shipments.
An AI-assisted intelligence report sent the US military to intercept a Chinese ship it wrongly believed carried nuclear components. Gary Marcus says he warned the Senate about exactly this.
A class-action complaint filed in San Francisco claims the four frontier AI developers agreed to slow product improvement, covering at least 80% of paid consumer subscriptions.
Anthropic has opened a wet lab in the San Francisco Bay Area where Claude models drive robots and lab equipment, with early tests showing a 50% hit rate in protein design.
Speaking at the University of Toronto, OpenAI co-founder and SSI CEO Ilya Sutskever said AI will keep improving until it handles all human work.
Anthropic and Accenture will each invest at least $1 billion over five years in independent evaluation of frontier AI, led by Accenture's Faculty unit.
Governor Abigail Spanberger signed Executive Order 22, barring NDAs for commercial data centers over 25 MW and creating an AI task force co-led by three state officials.
Moonshot AI's Kimi K3, the first open model with 2.8 trillion parameters, is now on Amazon Bedrock with a 1-million-token context window and prompt caching.
AI anxiety is becoming a “preference cascade”: polls show rising risk perceptions, while experts demand regulators, transparency and kill switches.
A survey puts the average AI researcher's estimate of human extinction risk at about 18%, while 93% of executives at a Yale event rejected the president's view.