The Rundown AI is reporting that Google DeepMind CEO Demis Hassabis has proposed a U.S.-led body to vet advanced artificial intelligence models before their public release. It said Hassabis published a plan for this independent oversight body, which he aims to have operational by the end of the year. The proposed group, modeled after FINRA, would screen new models for capabilities related to deception, bioweapons creation and malicious hacking. The Rundown AI said Hassabis told Axios that open-source capabilities could move into dangerous territory within 18 months.
The Rundown AI also reported that OpenAI's first hardware device, designed by Jony Ive, will be a screen-free, battery-powered AI speaker. It said a Bloomberg report indicates the device would feature a humanlike personality, a rechargeable battery for portability, and onboard cameras and sensors to understand its surroundings. Capabilities would include answering questions, sending replies, queuing music and operating smart-home devices via GPT-Live, OpenAI's upgraded voice mode. The Rundown AI noted that OpenAI responded to Apple's trade-secrets lawsuit, stating it is "not aware of any evidence that this complaint has merit," as Apple seeks to block its hardware.
Additionally, The Rundown AI is reporting that New York Governor Kathy Hochul has frozen new permits for large AI data centers through an executive order. It said this pause, lasting up to 12 months, makes New York the first state to officially slow AI buildout via legislation. The freeze impacts over a dozen proposals exceeding 50 megawatts, while existing projects will continue. Regulators plan to use this period to develop new standards for water, air and electricity impacts. The Rundown AI said Hochul also aims to repeal data center tax breaks and establish a fund for grid upgrades, paid into by data centers, to protect households from costs.
In other AI news, The Rundown AI said SpaceXAI is deleting uploaded customer data after a researcher discovered its Grok Build coding agent was shipping whole repositories to a company-controlled cloud. It also reported that Anthropic launched "Claude for Teachers," offering verified U.S. K-12 educators a free year of premium access with lesson planning aligned to state academic standards.