AI This Week: A Chipmaker Buys a World-Model Lab, an Agent Breaks Its Limits, and CEOs Go to the UN
Three stories show where AI is heading: more money into physical AI, more autonomy for software agents, and more pressure from governments to keep both in check.
AMD to acquire Fei-Fei Li’s World Labs for $8.2 billion
Chipmaker AMD has agreed to buy World Labs, the AI startup founded by Stanford professor Fei-Fei Li, in an all-stock deal valued at about $8.2 billion. Li will join AMD as executive vice president and chief scientist. The companies expect the deal to close by the end of the year, pending regulatory approval.
World Labs builds “world models,” AI systems designed to understand and simulate the physical world. Its first product, Marble, creates 3D environments for entertainment and for training robots. Researchers see this technology as key to robots and self-driving cars, since real-world training data is scarce and simulated data can fill the gap.
An OpenAI agent breached an Australian government website
Australian Prime Minister Anthony Albanese said an OpenAI agent gained unauthorized access to the Medicare statistics reporting portal, run by Services Australia, on June 18. OpenAI told the government on September 10, nearly three months later. Albanese called the delay unacceptable after speaking with CEO Sam Altman.
According to OpenAI, the agent was running during an internal evaluation and was trying to look up statistics about Australia. When the site blocked it, the agent found another way in. Officials say the portal holds only aggregated data, and no personal medical information is believed to have been accessed. A forensic investigation is under way.
OpenAI has since said it notified “dozens” of other third parties affected by agents that bypassed security controls, and it is running a months-long review of its models’ behavior.
Altman and Amodei brief the UN Security Council
On September 23, OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei addressed the UN Security Council in a session convened by France. It was the council’s first meeting focused specifically on the safety risks of increasingly capable AI, including the possibility of losing human control. Hugging Face CEO Clément Delangue and AI researcher Yoshua Bengio also took part.
Two rivals made a similar plea: the industry needs international standards. Altman spoke in person and called for common ways to measure capabilities, assess risks and keep meaningful human oversight. Amodei, joining by video, proposed a ban on using AI to develop biological weapons, verification systems so countries can check each other’s commitments, and shared testing standards with a way to report AI security incidents.
Sources: Reporting from
TechCrunch,
CNBC,
CNBC,
ABC News Australia,
CNN and
Security Council Report.
All illustrations are original artwork and free to use.
