Anthropic resumes charging for safeguard-blocked requests in three categories

The policy applies to blocks in biology, distillation attacks, and frontier LLM development to counter coordinated system attacks.

• Classifiers for these billable blocks are tuned to a false positive rate under 0.1%.
• In recent testing across Claude Code, Claude.ai, and Cowork, 99.7% of accounts did not encounter billable blocks.
• Developers can flag incorrect blocks using the /feedback command in Claude Code.

Users running requests flagged for biology, distillation, or frontier LLM development will now incur costs even when requests are blocked before generation.

Sources

• Anthropic: Anthropic resumes charging for safeguard-blocked requests in three categories

Read this as text

Back to the AI news