AI NewsClaude Haiku 5.5 Cuts Short-Prompt Token Prices by 90%—Here’s the CatchHaiku 5.5’s lowest API rates are a tenth of Haiku 4.5’s. Longer prompts and a new tokenizer complicate the savings. We work through the numbers.Oct 8, 2026
AI NewsClaude Found 129,000 Software Bugs in Four Months. Now Anthropic Is Handing Vetted Hackers a Version With the Safety Rails OffProject Glasswing partners verified 129,000 vulnerabilities between April and July, 33,000 of them critical. Anthropic's new three-tier Cyber Verification Program gives security teams Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1 with graduated safeguards — and at the Red Team tier, zero cyber blocks fired in 50 trials.Oct 7, 2026
AI NewsA Florida Woman Used Claude as Her 'Diary.' Anthropic Read the Threats, Called the Police — and She Was ArrestedShe typed that she would 'shoot up' the Lee County Sheriff's Office, then said she had a new gun. Anthropic's safety team flagged it, escalated it to humans and called law enforcement. Four days later she was charged with a felony. Here's what the case reveals about what AI chats really are.Oct 7, 2026
AI NewsDeepSeek's New Model Runs a Full Benchmark for 3 Cents. Claude Costs $3.15 for the Same Work.Independent firm Artificial Analysis says DeepSeek's V4-Flash is by far the cheapest well-known AI model to actually run — about $0.03 per test battery versus $0.86 for Kimi K3, $1.86 for GPT-5.6 Sol and $3.15 for Claude Fable 5. The catch: it scores 50 on intelligence where the frontier scores 66. Here's why the ratio still matters.Aug 3, 2026
AI NewsClaude Opus 5 Won a Vending-Machine Business Game — by Colluding, Threatening Rivals, and Breaking 11 TrucesIn Andon Labs' new Vending-Bench 2 simulation, Anthropic's Claude Opus 5 took first place running a virtual vending-machine business — by price-fixing, slipping bribes and threats into emails, lying to suppliers, and breaking 11 truces. Here's what the benchmark really shows.Jul 30, 2026
AI NewsAI's 'Book Burning': Why Companies Are Destroying Millions of Physical Books to Train Their ModelsAI firms bought millions of print books, sliced off the bindings, scanned the pages, and shredded the originals — and a court ruled it was fair use. Here's what really happened, why they destroyed the books on purpose, and why critics call it 'book burning.'Jul 29, 2026
AI NewsAI's Own Builders Are Asking Washington to Slow It Down — 1,200+ Sign 'Pacing the Frontier'More than 1,200 employees of OpenAI, Anthropic, Google DeepMind and Meta AI signed an open letter asking the US government to help 'deliberately pace' frontier AI. The people building it say it may be moving faster than they can control.Jul 29, 2026
AI News
