Anthropic’s Fable Under Fire
- Anthropic’s Fable has been criticized for its guardrails by cybersecurity researchers.
- The guardrails can be bypassed by adding specific prompts to malware.
- Anthropic has apologized and announced changes to the safeguards.
The Buzz Score
The Internet’s Verdict: 70% Hyped, 30% Skeptical
Expert Opinions
Cybersecurity researchers have expressed concerns about the effectiveness of Anthropic’s Fable guardrails.
Malware authors are pretty excited about guard-rails. you can add prompts to your malware to get LLM scanners to hit guard-rails and stop their runs.
Some researchers have also criticized the lack of transparency in the guardrails.
The strangest part is that it won’t just reject ML research, which I can understand, it will sabotage it silently by using a worse model without revealing it is doing so.
Anthropic has announced changes to the safeguards in response to the criticism.
We’re changing Fable 5’s safeguards for frontier LLM development to make them visible. We made the wrong tradeoff and we apologize for not getting the balance right.
Focus Keyword: Fable Security
