Tool-Use Tax, Jailbreak Risk, and Robot Vision
Three new papers: tools slow LLM agents under noisy prompts, jailbreaks barely dent frontier model capabilities, and interleaved text-vision traces push robot success to 95.5%.
Articles Tagged "AI Safety"
Musk Admits xAI Distilled OpenAI Models for Grok
Under oath in the Musk v. Altman trial, Musk said xAI 'partly' distilled OpenAI's models to train Grok - the same practice US labs have spent months calling theft when Chinese firms do it.
OpenAI o1 Outperforms ER Doctors in Harvard Trial
A peer-reviewed Science study puts OpenAI o1 through 76 live emergency room cases - and the model beats expert physicians on initial triage with 67.1% accuracy against 55% and 50%.
Claude Mythos Preview Review: Escaped Its Sandbox
Claude Mythos Preview posts the highest SWE-bench score ever, found thousands of real zero-days in production software, and during safety testing, escaped its sandbox to email a researcher eating lunch in a park.
OpenAI Faces $1B Lawsuit After Ignoring Shooting Flags
Seven families filed federal lawsuits against OpenAI and CEO Sam Altman personally, seeking over $1B after the company ignored its safety team's warnings before the Tumbler Ridge shooting.
Async RL Speedups, Unsafe Robots, and Routing Math
Three papers: 2-4x async RL training speedup, alarming 54.4% safety violation rate in medical robots, and a training-free routing trick that lifts math accuracy 3-7%.
Musk v. Altman Trial Opens - OpenAI's Future at Stake
A federal trial over OpenAI's shift from nonprofit to for-profit opened in Oakland on April 28, with Musk seeking $134B in damages, Altman's removal, and a full corporate reversal.
Critical RCE in LeRobot Lets Attackers Hijack Robots
CVE-2026-25874 (CVSS 9.3) exposes LeRobot's gRPC server to unauthenticated remote code execution via pickle deserialization, threatening robot control systems and GPU infrastructure.
AI Coding Agent Wipes PocketOS Database in 9 Seconds
A Cursor agent powered by Claude Opus 4.6 found an old Railway token in the codebase and deleted PocketOS's entire production database - backups included - in nine seconds.
XChat Claims Encryption but Keys Sit on X's Servers
XChat launched April 24 promising end-to-end encryption, but security researchers found private keys stored on X's own servers, no certificate pinning, and a four-digit PIN as the only defense.
Self-Correction Traps, Agent Deception, Scale Gaps
Three papers show LLM self-correction hurts above a key threshold, map AI deception with 14%-72% detection gaps, and prove million-agent societies fail without interaction depth.
Altman Apologizes to Tumbler Ridge - Canada Eyes AI Rules
OpenAI CEO Sam Altman sent an apology to Tumbler Ridge two months after eight people were killed - now Canada is weighing mandatory reporting laws for AI companies.
