Anthropic puts model behavior at the center with Claude Opus 4.7 honesty and anti-sycophancy claims
Anthropic says Claude Opus 4.7 improves honesty to 92% and reduces sycophancy, highlighting a new front in enterprise AI model competition.
AI Accuracy
AI Fact-Checking Remains Unreliable, WIRED Analysis Finds
A WIRED fact-checker tests AI systems and highlights persistent reliability problems in automated fact-checking workflows.
Google's AI Overview Struggles with Basic Calendar Math, Confuses Year 2027
Futurism reports Google's AI Overview feature continues to hallucinate, incorrectly stating that 2027 is not next year, highlighting persistent accuracy issues in AI systems.
Featured
AI Accuracy
Latest News and Analysis in AI Accuracy