GPT-6 is here (with critical security risks)

Semafor

Senate lawmakers are making progress on bipartisan AI safety legislation targeting immediate risks like model approvals and harm alerts, with committee Chair Ted Cruz expressing hope for a markup vote before midterm elections, though significant political and industry hurdles remain.

Bloomberg

Apollo is in talks to increase its loan facility to SoftBank from $5.4B to $9B to finance the Japanese conglomerate's OpenAI investments, signaling massive capital flows into leading AI companies and the critical role of alternative financing in the AI race.

Anthropic

Anthropic is consolidating its Cowork feature into the main Claude interface and launching Docs and Slides capabilities in beta, allowing the AI to autonomously handle complex multi-format projects within a single conversation without users needing to choose between different tools.

OpenAI

OpenAI's GPT-6 Astra represents a major leap in AI capabilities with near-perfect benchmark scores and advanced computer use abilities, but its breakthrough cybersecurity skills—including discovering zero-day exploits—have necessitated unprecedented safeguards and restricted access for defensive security work.

Novo Nordisk

Novo Nordisk is partnering with Anthropic to use Claude Science and frontier AI models for drug discovery and software engineering, accelerating R&D timelines as part of its ambition to become the world's most AI-driven healthcare company.