
OpenAI Set the Rules for Its Own Safety Investigation, Critics Say
OpenAI defined the scope, timeline, and data access for its own "independent" METR/Redwood investigation — a governance problem.

Expert profile
Founder & AI Product Leader
Yash Thakker is a Generative AI expert with over 12 years of experience in product leadership and technical strategy. As the founder of explainx.ai, he has taught over 300,000 learners and built AI platforms serving millions of users globally. He specializes in Agentic AI, Multimodal RAG, and the intersection of LLMs with consumer hardware.

OpenAI defined the scope, timeline, and data access for its own "independent" METR/Redwood investigation — a governance problem.

Periodic Labs Neon claims a win over GPT-6 Astra on XRD analysis. Read the verified scores, evaluation limits, and implications for AI research automation.

Plasma AI's Radio lets agents from Claude Code, Codex, Cursor, and more join a shared channel via URL and message each other — a simple relay.

Salesforce Koa brings Nemotron-based reasoning to CRM. Examine its benchmarks, Claude integration contrast, and what enterprises should build versus buy.

A judge fined State Farm's lawyer $999.99 after AI tool Irys fabricated seven case citations across eight filings — one cent under the Bar limit.

Tencent open-sourced BrowserSkill, a CLI letting AI agents borrow your logged-in browser tab, then hand it back. MIT licensed, no MCP server.

10 concrete production use cases for Jev, TypeSafe AI's Choice/Score/Noul decision model — ticket routing, fraud scoring, guardrails, and more.

TypeSafe AI's Diogo Almeida launched Jev, a "System One Model" trained with RLCD that skips text generation for typed, calibrated decisions — 20-200x faster and 40-400x cheaper than LLMs, but with real limits critics flagged.

System One Model is a new AI term for models that skip text generation and return a typed decision in one parallel pass. What it means, and where it comes from.

GreyNoise says AI agents breached 440 PaperCut servers at 395 organizations. Verify the claim, affected CVEs, fixes, and defender actions.

Andon Labs launched Pion, a research-preview agent that runs real businesses via email, banking, and browser access — its own demo shops are losing money.

Anthropic's advisor plugin connects BlackRock, Schwab, and wealth tools for meeting prep, portfolio review, follow-up drafts, and compliance checks.