Chinese AI Firms Accused of Copying US Models as OpenAI Adds Doomer Paul Christiano to Board
Six Chinese AI firms have been accused of aggressively copying US frontier models, prompting the US government to urge domestic AI companies to identify and then secretly switch Chinese users to less-capable versions, according to reports. The tactic represents a significant escalation in the AI Cold War between Washington and Beijing, targeting firms attempting to replicate products from OpenAI, Anthropic, and other leading developers.
OpenAI added prominent AI safety researcher Paul Christiano to its board of directors, deepening the company's commitment to alignment oversight at a time when competitors are racing to ship increasingly capable systems, according to Bloomberg. Christiano, an influential figure focused on AI alignment, will join the OpenAI Foundation board as the company navigates mounting regulatory pressure across the US and Europe.
An Anthropic researcher resigned with a stark warning that self-improving AI could "kill us all," telling colleagues that the team earnestly believes artificial intelligence poses an existential threat to humanity, per reports. The departure highlights growing internal tensions at leading AI labs between safety-minded researchers and teams pushing to accelerate development timelines.
AI research startup Listen Labs walked away from a signed $1.5 billion Series C term sheet from Menlo Ventures as it pursues acquisition talks with Salesforce, sources told The Information. The abrupt decision to scrub the massive funding round underscores how strategic M&A opportunities are increasingly disrupting traditional venture financing paths for promising AI startups.
Suno released its v6 AI music model, the first built with support from the record industry after months of licensing negotiations, according to The Verge. Suno's Jack Brody said v6 was "trained from the ground up with a new set of data" that excludes the unlicensed material that fueled earlier versions, a shift prompted by ongoing litigation from major labels.
Anthropic granted the EU cybersecurity agency ENISA testing access to its Mythos 5 model after months of talks that began in late May 2025, though ENISA still lacks access to the newer Mythos 5.1, reports Bloomberg. The partial access represents incremental progress in Anthropic's relationship with European regulators, who have demanded greater transparency into frontier model capabilities.
Sources: