Anthropic AI Updates: June 10, 2026

1. Anthropic Releases Claude Fable 5 and Restricted Claude Mythos 5

Anthropic. Anthropic announced Claude Fable 5, a new Mythos-class model with built-in safeguards that it describes as state-of-the-art on nearly all tested capability benchmarks, alongside Claude Mythos 5, the same underlying model with safety restrictions removed for authorized users such as cybersecurity professionals and selected researchers. The company cites a Stripe codebase migration of 50 million lines completed in one day, top scores on finance and trading evaluations, and Mythos 5 generating molecular biology hypotheses that scientists preferred in 80% of head-to-head comparisons. Fable 5 is available globally via the API and subscription plans at $10 per million input tokens and $50 per million output tokens, while Mythos 5 is limited to Project Glasswing partners and selected biology researchers. Source

2. Claude Managed Agents Adds Scheduled Runs and Environment Variable Vaults

Anthropic. Anthropic added scheduling to Claude Managed Agents, letting agents run tasks automatically on cron schedules without users building or hosting their own scheduler, with controls to pause, resume, archive, or trigger additional runs manually. The release also introduces environment variable vaults so that CLIs and authenticated tools access credentials securely: the sandbox holds only a placeholder while the real credential attaches at the network boundary and is restricted to approved domains. Anthropic cites Rakuten, Actively AI, and Ando using the features for automated reporting, cross-account search, and workflow management, along with vault-authenticated browser access through the Browserbase and KERNEL CLIs. Source