Anthropic's Cyber Classifier Silently Overrides Specified Model ID
alexcovo_eth · x · 2026-08-11
During a read-only adversarial review, the author explicitly specified the claude-fable-5 model ID. However, Anthropic's cyber classifier intercepted the request in less than a second and silently switched the session to Opus 4.8.
- Behavioral Details: The raw stream shows only one event from Fable 5 (containing only thinking). The subsequent 50 events, including file reading and the final verdict, were entirely generated by Opus 4.8.
- Data Evidence: The final modelUsage data lists only Opus 4.8 and Haiku, completely ignoring the user-specified model.
This indicates that when a request is flagged, the specified model ID is treated as optional and overridden by a fallback model.
More from Models
- User questions why OpenAI doesn't RL against specific 'slop' in model outputs — kalomaze · 2026-08-12
- Grok Build v1.0.1 Adds Bounded Subagents and Safer MCP Controls as Musk Seeks Feedback — elonmusk · 2026-08-12
- Google Exec Touts Gemini API's Generous Permanent Free Tier — sunjiao123sun_ · 2026-08-12
- Grok Users Report Heavy Censorship, Speculate X IPO Compliance — DevDminGod · 2026-08-12
- Microsoft's New Code Model Boosts Efficiency 25% at Quarter of the Cost — mustafasuleyman · 2026-08-12
- Users Report Grok Unreasonably Refusing Cutting-Edge Science Equations — Promptmethus · 2026-08-12