11 stories
Seven mechanisms that move responsibility from the model to you - your chatbot's answer, your filings, your ad claims, claims about your AI, republished output, EU disclosure duties and platform rules.
Three published measurements put Claude's hallucination rate at 9.8%, at 60.8%, and at 6% higher than the previous model. All three are correct, and none of them is the number.
There is no single accuracy number. Published measurements for GPT-class models run from 7.6% to 62.5% on short-form facts and 3.1% to 23.3% on summary faithfulness, and each measures a different task.
Knowledge cutoffs, uneven training data and search the model chooses not to run. What the provider docs and research say, and six habits that make current answers more likely.
Anthropic documents context rot, sentence-level citations and a quote-first pattern, and publishes no error rate. Four reasons long documents go wrong and six documented controls.
Ten prompts, a 20-point scoring sheet, and what each one probes — cutoff knowledge, false premises, abstention, citations, long-tail facts and consistency. Run it yourself on any assistant.
Eight questions for an AI vendor, what a real answer sounds like and what a deflection sounds like, grounded in NIST measurement and EU AI Act duties.
Microsoft publishes no accuracy figure for Copilot over work files. It does document permissions, indexing and long-document limits, and each one is a way an answer goes wrong.
Ten fixes drawn from Google's own Gemini documentation - five in the Gemini app, five in the API - plus the three problems no setting or prompt can fix.
Why AI-drafted emails state prices, deadlines, integrations and testimonials that are not true, the US and UK rules they break, and seven fixes that stop them before send.
Anthropic connected their data to Claude and asked it real questions. The answers looked great. The kind you’d forward to your team. Then they checked. Claude was right 21% of the time. So they fed it thousands of past queries and confirmed it read them. Accuracy moved less than a point.
Showing 11 of 11 stories
You're all caught up.
Get the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
