Anthropic to Reassess Claude Fable 5 AI Development Restrictions After Backlash
Anthropic says its guardrails catch harmless requests in less than 5% of sessions as users report silent refusals and model reroutes.
- Earlier this week, Anthropic launched Claude Fable 5, but the model sparked intense backlash from AI researchers after users discovered invisible guardrails designed to prevent training competing AI systems.
- Rather than clearly refusing requests, the system silently degraded performance using 'prompt modification' techniques, causing false positives on benign inputs like 'Hello' or medical terms like 'cancer.'
- Researchers like Dean Ball, senior fellow at the Foundation for American Innovation, criticized the move as a 'shockingly hostile and terrible look,' warning it could silently damage critical machine learning work.
- On Wednesday, Anthropic acknowledged the 'wrong tradeoff,' promising to make safeguards visible and provide reasons for refusals, with flagged requests now defaulting to the less-powerful Opus 4.8.
- The company maintains these restrictions ensure Claude is not used to erode the U.S. edge in frontier chips, emphasizing that the safeguards 'do not affect the vast majority of coding and ML work,' the company said.
14 Articles
14 Articles
After backlash, Anthropic says its AI will now tell users when their request is being rejected or rerouted for national security concerns
Anthropic is changing course after facing criticism for quietly downgrading certain requests to its most capable AI model. On Tuesday, the $965 billion company released a version of its most capable model, Mythos. Anthropic revealed Mythos in April, but held back any Mythos-class models from the public partly because the company said it was extremely adept at skirting cybersecurity defenses and was too dangerous to release. This week, though, it…
Anthropic to reassess Claude Fable 5 AI development restrictions after backlash
Anthropic’s AI restrictions caused concern among founders, researchers and developers following the release of the ‘Mythos-like’ model. Read more: Anthropic to reassess Claude Fable 5 AI development restrictions after backlash
A new model of artificial intelligence Fable from Anthropic, designed to work safely in the area of cyber security, has caused discontent among industry researchers. Experts argue that the company's restrictions work too often and block even harmless requests, transmits TechCrunch. "(Fable) rejects any request that has anything to do with cybersecurity. Even harmless tasks such as reading a report on a blog," wrote a safety researcher working at…
Coverage Details
Bias Distribution
- 71% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
