Claude Fable 5 reroutes high-risk queries
Anthropic has launched Claude Fable 5, its first publicly available 'Mythos' class model, which silently reroutes high-risk queries to Claude Opus 4.8 to manage safety. While general users can access the Fable 5 tier, the unrestricted Mythos 5 model remains gated for vetted security organizations.
Anthropic's silent routing strategy shows that frontier models are becoming too powerful to align at the training level alone, forcing a pivot to runtime traffic redirection.
- –Silently rerouting high-risk queries to an older model (Opus 4.8) is a pragmatic safety compromise, but it undermines developer transparency and predictability.
- –Keeping the unrestricted version (Mythos 5) behind closed doors highlights that the underlying capabilities of this model class are still deemed highly dangerous.
- –A mandatory 30-day data retention period reveals that post-deployment monitoring is now a critical layer of defense for agentic AI.
DISCOVERED
1h ago
2026-06-11
PUBLISHED
2h ago
2026-06-11
RELEVANCE
AUTHOR
aieonme