OpenAI Launches Astra for Law
The system combines a frontier model with a large US legal index, specialized instructions and controls for confidential work.
The system combines a frontier model with a large US legal index, specialized instructions and controls for confidential work.
A preprint studies how harmful behavior can move through multi-agent handoffs and separates mutation, transmission and recovery.
A new study argues that refusal behavior can collapse at the start of reasoning and tests a learned safety signal at that boundary.
A proposed monitor checks an agent’s complete action history because individually permitted steps can combine into a prohibited outcome.
ByteDance retains control of its AI drug-discovery spinout after an external round that reportedly values the company at $1.5 billion.
The enterprise-AI providers plan a transatlantic company with dual headquarters and new backing from Schwarz Group.
A standing process will publish unexpected model behavior sooner, including cases that remain unexplained or only partly mitigated.
Cowork and chat now share one interface, while new document and slide tools move Claude closer to an end-to-end work surface.
Never Give Up keeps sampling a problem until it finds a correct answer, shifting reinforcement-learning compute away from examples the model already solves.
The voluntary framework recommends at least $1 million in local investment for every megawatt of utility demand from a proposed project.
Mark Zuckerberg argues that competition, liability and independent evaluation give each laboratory reasons to act safely without collective limits.
Reuters says the agreement covers a planned 2.16-gigawatt campus near Brisbane. Anthropic and developer Zerra DC declined to confirm the deal.