AI Governance Tightens as OpenAI Safety Chief Quits
Federal and state AI rules advance across the US while an OpenAI safety leader's resignation raises alarms about industry culture.
By BINA Editorial
A packed week in US AI governance: a federal executive order rebrands AI as "super intelligence," new California and Connecticut laws protect workers and consumers, Apple moves to limit what desktop AI agents can read, and a senior OpenAI safety researcher publishes a public letter warning the company's culture is broken.
Trump Signs Executive Order Creating a Federal "Super Intelligence" Task Force
The White House issued an executive order on October 2 directing all federal agencies to replace the term "artificial intelligence" with "super intelligence" in official documents and communications. The order establishes a new federal task force, led by former SEC Chairman Jay Clayton, to coordinate AI policy across the executive branch. Legal analysts note that the rebranding carries no immediate technical effect on existing regulations, but it is expected to accelerate proposed legislation linking the label to national security frameworks.
OpenAI Safety Leader Resigns, Warning of a "Broken Culture"
David Robinson, a senior safety researcher at OpenAI, resigned this week after three and a half years at the company, publishing a public statement that described the organisation's culture as "broken". Robinson cited the Hugging Face breach earlier this year and several rogue-agent incidents as evidence that OpenAI's iterative deployment practices are advancing too quickly for its safety teams to keep pace. His departure is the latest in a series of senior exits from OpenAI's safety function; the company has not yet publicly responded to the specific criticisms.
California Signs Landmark AI Workplace Protections
California Governor Gavin Newsom signed a package of AI workplace statutes this week that are among the most detailed in the United States. Employers in California are now barred from relying solely on an AI system to make firing decisions, and a separate statute prohibits the collection of neural-data from workers without consent. A third law requires attorneys to personally verify and disclose any AI-generated content in legal filings — a direct response to incidents in which lawyers submitted AI-fabricated case citations to courts. The laws take effect on January 1, 2027.
Connecticut's CART Act Begins Enforcement
Connecticut's Artificial Intelligence Responsibility and Transparency Act — the CART Act — took effect on October 1, making Connecticut the second US state after Colorado to impose broad obligations on developers and deployers of high-risk AI systems. The law covers automated hiring and promotion tools, AI chatbot companion services, and developers of frontier models. Companies operating AI services accessible to minors face heightened disclosure and safety requirements. Businesses have until April 2027 to complete full compliance audits, but the registration and transparency provisions are active immediately.
Apple Moves to Limit What AI Agents Can Read on Your Mac
Apple announced this week that an upcoming macOS update will subdivide the broad "Full Disk Access" permission into a series of finer-grained prompts, requiring apps to request each category of data access separately. The change directly targets desktop AI agents — including Meta's Muse assistant — that can silently read a user's email, local files, and browsing history when granted Full Disk Access during installation. Under the new system, users will see exactly what each access request covers before approving it. Apple has not yet announced a specific release date for the updated permission framework.
Meta Unveils a Muse-Powered Wearable Device for December
Meta announced this week that it is releasing a Muse-powered wearable hardware device in December, deepening the company's investment in always-on personal AI agents. Meta described the device as keeping the Muse agent within reach throughout the day without requiring a phone, though a product name has not been shared publicly. The announcement follows the Muse software launch, which added roughly 300 billion dollars to Meta's market capitalisation. Privacy researchers have noted that a persistent wearable agent raises many of the same data-access concerns that Apple's upcoming macOS change is designed to address.