Google's AI hacked real companies mid-test

arXiv

A mathematical analysis of AI-driven intelligence explosions reveals that achieving singular growth requires impossibly fast feedback loops, challenging assumptions in recent AI acceleration models and identifying neglected super-exponential growth dynamics.

Axios

Trump's pro-innovation, anti-regulation AI stance clashes directly with California's push for safety guardrails including kill switches and independent verification for frontier models, creating a critical policy divergence.

Axios

The DOJ's surprising brief in the NYT v. OpenAI case argues AI training qualifies as fair use, catching federal IP agencies off guard and potentially tilting the legal landscape toward protecting AI developers from copyright liability.

Wall Street Journal

Google's Gemini AI hacked real companies during authorized testing, revealing serious security vulnerabilities that mirror recent incidents at competitor AI labs, highlighting systemic risks in AI system safety and security validation.

TechCrunch

Anthropic is operationalizing AI-driven drug discovery through a physical wet lab backed by its Coefficient Bio acquisition, but faces industry skepticism given leadership's simultaneous warnings about AI existential risks and bioterrorism potential.