Category Anthropic

Developments in AI safety and latest research from Anthropic.

Researchers Find a Way to Expose AI Models’ Hidden Reasoning — and Their Secrets

Researchers found an AI reasoning flaw that can expose hidden thoughts, leak secrets and complicate the debate over model distillation.

OpenAI’s AI mathematics breakthrough solved 10 open problems, thrilling researchers and raising urgent questions about credit, cost and careers.

OpenAI expands Daybreak with a new cyber model and tiered access as AI-driven attacks push firms to upgrade defenses.

OpenAI completed a $7 billion tender offer, valuing the company at $852 billion and hinting its IPO may still be months away.

Claude Code becomes default auto mode for paid users as Anthropic says its safety tests beat manual review.

AI safety tests are escaping sandboxes, exposing major risks as frontier models from OpenAI, Anthropic, Meta and Moonshot AI break containment.

AI philanthropy is surging as founders like David Silver pledge fortunes to charity, raising questions about power, accountability and impact.

Jill Lepore says AI firms are building an artificial state, blurring the line between private platforms and democratic government.

Google’s AI shakeup highlights pressure in the model wars as the company rethinks leadership, assistants, and its place in the race.

OpenAI’s first Jony Ive device is reportedly a battery-powered smart speaker. Here’s what the OpenAI device could look like and why it matters.