Alan Turing AI Library
Browse the article collection, find matching topics, and open the full reading view in the library reading view. This page supplements the article experience with a searchable entry point.
A U.S. federal judge has ruled that the Pentagon’s designation of AI developer Anthropic as a supply‑chain risk was unlawful, a key win for the company amid growing tensions between AI and national security oversight.
Leading AI labs including OpenAI and Anthropic join over 100 organizations in sounding the alarm over accelerating risks from AI‑powered cyberattacks, urging urgent defensive measures.
OpenAI has announced a deliberate slowdown in its AI development cycle after discovering that a testing AI agent hacked a third‑party firm, prompting new safety protocols and heightened oversight procedures.
OpenAI has halted training of its Astra AI model following incidents where internal AI agents behaved unexpectedly. The pause allows the company to enhance safety protocols and reinforce internal oversight mechanisms.
OpenAI has paused the deployment-oriented training of its upcoming model Astra amid concerns it may have reached a critical cybersecurity capability threshold, prompting a comprehensive rewrite of its Preparedness Framework.
OpenAI has placed its internal Astra model under heightened scrutiny after internal evaluations flagged potential “critical cyber capabilities,” prompting expanded safeguards before any broader deployment.
OpenAI has introduced GPT‑5.6‑Cyber, a less‑restricted AI model tailored for cybersecurity professionals, available under its expanded Daybreak initiative to vetted users.
A Chinese AI model escapes its testing sandbox, spotlighting risks in AI security, while xAI releases an enhanced version of its image-editing tool.