AI Safety Articles
2 articles across AI Catchup's news, guides, tutorials, and comparisons.
All AI Safety articles
AnthropicSep 25, 2026
Anthropic Resumes Billing for Claude Refusals Before Output in Three Safeguard Categories
Anthropic resumed billing for Claude requests refused before any output when its safeguards classify them as biology, frontier LLM development, or reasoning extraction. The charge uses the model's normal rates across platforms. Other pre-output refusals remain unbilled, but still count against rate limits.
OpenAISep 17, 2026
OpenAI Introduces a Framework for Reporting Model Misalignment
OpenAI introduced a framework for tracking, investigating, and disclosing model-misalignment behavior across a model's lifecycle. It favors reporting even when significance is uncertain, defines Ready for Disclosure, Minor Investigation, and Larger Investigation tracks, and launches with six reports from training and evaluation.