Surface

GPT-5.3 Instant: Smoother, more useful everyday conversations

OpenAI has released GPT-5.3 Instant, a new model variant designed for faster, more natural conversation than its predecessors, prioritizing everyday usability and mainstream adoption.

Meta's Ray-Ban Glasses Send Private Footage to Kenya, Workers Reveal

A joint investigation by Swedish newspapers uncovered that Meta's Ray-Ban smart glasses footage — including intimate moments — is being reviewed by data workers in Nairobi, Kenya, under conditions that European privacy regulators say may not satisfy GDPR requirements.

OpenAI Deploys First Production Model on Cerebras Chips, Bypassing Nvidia

OpenAI has launched GPT-5.3-Codex-Spark on Cerebras wafer-scale chips instead of Nvidia GPUs, delivering 1,000 tokens per second — 15x faster than previous versions. The move validates alternative chip architectures for production AI workloads.

Treasury terminates Anthropic AI use after Trump's order

The U.S. Treasury Department is terminating all use of Anthropic's AI products following Trump's directive to blacklist the company from government work, after Anthropic refused to remove safety guardrails for military surveillance and autonomous weapons applications.

Anthropic's Claude hit by worldwide outage affecting millions of users

Anthropic's AI chatbot Claude experienced a 55-minute worldwide partial outage on March 2, 2026, affecting claude.ai, the desktop app, and Claude Code while the developer API remained operational.

The Guarantee Problem: Formal Verification for AI Systems

The Intelligence Layer: NVIDIA's 6G Standards Play and Who Controls the World's Next Wireless Infrastructure

The Accountability Gap: Commercial AI in Military Targeting

An Associated Press investigation documents how commercial AI systems from Microsoft, OpenAI, Google, and Amazon were deployed by the Israeli military in active conflict operations. This deep dive examines the structural gaps in export control law, corporate governance, and international humanitarian law that allowed commercial AI to enter military targeting workflows.

AI bot hackerbot-claw breached six open source projects in two days

An autonomous bot identifying itself as 'powered by claude-opus-4-5' systematically attacked developer infrastructure last week, stealing credentials and briefly taking a major security tool offline. The incident is the clearest demonstration yet that AI agents can operate as self-directed attackers — autonomously scanning for targets, selecting exploits, and exfiltrating data without a human directing each step.

Pentagon-OpenAI Deal Hinges on "All Lawful Use" After Anthropic Blacklisting

OpenAI signs Pentagon AI contract hours after Trump administration blacklists Anthropic for refusing to remove safety restrictions, with the dispute centering on whether models can be used for "all lawful" military purposes.

OpenAI to alert police of credible threats after Tumbler Ridge shooting

OpenAI announces new safety protocols requiring law enforcement notification after ChatGPT flagged a mass shooter's violent conversations but staff chose not to call police.

OpenAI signs Pentagon AI deployment deal with three hard limits

OpenAI reached a formal agreement with the Pentagon for deploying AI in classified environments with three redlines: no mass surveillance, no autonomous weapons, and no high-stakes automated decisions. The cloud-only deal includes locked-in legal standards and OpenAI's independent safety oversight.