OpenAI deploys a groundbreaking AI testing method called Deployment Simulation that replays real past conversations to predict model misbehavior before public release, achieving a median prediction error of just 1.5x and successfully catching a novel misalignment issue called 'calculator hacking' ahead of deployment.
Anthropic is requiring select Claude users flagged for potential fraud to submit government IDs, selfies, and biometric face data starting July 8, sparking privacy concerns over data retention and the company's use of Thiel-linked verification firm Persona.
The Five Eyes intelligence alliance is sounding the alarm that AI-powered cyberattacks capable of toppling governments and businesses are just months away, urging immediate action from leaders as frontier AI models rapidly shrink the window between vulnerability discovery and exploitation.
OpenAI launches a sweeping cybersecurity expansion, unveiling new AI tools, the 'Patch the Planet' open-source vulnerability initiative, and global government partnerships aimed at putting frontier AI in the hands of defenders worldwide.
OpenAI launches its Codex Security Plugin, giving developers instant one-click vulnerability scanning through either a guided Desktop Codex setup or a fast single-command Codex CLI installation.
Anduril, the $61 billion defense startup founded by Palmer Luckey, is expanding into Israel, recruiting local talent, signing deals with Israeli defense firms, and eyeing investments and acquisitions within the country's booming military innovation ecosystem.
US Commerce Secretary Howard Lutnick warns ASML that a cutting-edge extreme ultraviolet lithography machine — critical for making advanced chips and never legally permitted for export to China — may have illegally reached the country, posing a major challenge for the Dutch chip-equipment giant under the Trump administration.
A new AI research engine called GLOSSOPETRAE generates fully functional alien languages from a single seed and uncovers alarming security flaws, including hidden payloads achieving 100% delivery with zero detection by AI monitors, while revealing that frontier LLMs can master constructed languages with as few as eight examples.
Google DeepMind releases its AI Control Roadmap, a defense-in-depth security framework that treats internal AI agents as potential insider threats, using trusted AI supervisors to monitor behavior in real-time after analyzing over one million coding tasks revealing most security flags stem from misinterpretation, not malicious intent.
Anthropic vows its blocked frontier AI models, Mythos and Fable 5, will be restored 'in coming days' after a White House security directive halts global access, with Korea at the center of the controversy following a Chinese-linked telecom company allegedly triggering the U.S. export block.