Holy smokes! Anthropic models will deliberately disallow tasks that are identified as self-improving for other AI models.
12:06 AM · Jun 10, 2026 · 812 Views
Anthropic reportedly restricts machine learning research queries and blocks tasks classified as self-improving for other models
Story Overview
Anthropic's freshly released Claude Fable 5 and Mythos 5 models arrive with moderation layers that refuse or quietly downgrade responses on machine learning tasks the system flags as self-improving, including GPU inference work and advanced programming queries from paid users.
1111842612.5K
Developer Impact
Performance hits paying users first
Observers tracking the new models report deliberate quality drops on technical workloads, turning what should be top-tier assistance into slower or incomplete answers for exactly the developers who pay for priority access.
Open Question
Scope of the blocks stays fuzzy
The system card flags limits on frontier-model research, yet exact triggers, how widely they apply beyond high-volume accounts, and whether the degradation is permanent remain unclear from current reports.
Sentiment
Negative users criticize Anthropic for blocking ML and GPU research in its models over LLM competition fears, viewing the restrictions as excessive control and a sign of weakness.
Pos
0.0%
Neg
100.0%
6 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS845BOOKMARKS6LIKES16
Ross Taylor@rosstaylor90
Hopefully it is obvious now that if your country’s sovereign AI strategy does not concentrate on the model layer, it is going to have a hard time.
All advanced technology is now downstream of model intelligence.
1hViews 845Likes 16Bookmarks 6
stanlΞy@stanthepixelman
@GergelyOrosz nerfing your own model in case the customer might be a competitor is a wild thing to choose to ship
1hViews 79Likes 2Bookmarks 1
Krystian Laskowski@KrystianLaskow1
@GergelyOrosz Proof they have no moat. I don't see Anthropic surviving next 20 years, at least not as a top frontier coding model provider.
1hViews 38Likes 1
Sid Sanyal@siddsanyal
@GergelyOrosz And Anthropic trained its models on data available on the web
1hViews 61
Daniel James Tronca@danieltjamess
@GergelyOrosz I was switched to Opus 4.8 even while I was working on health algorithms for a fitness app 🤷🏼♂️
1hViews 27
Bruno Dias@BrunoaaDias
@GergelyOrosz I think OpenAI do the same. I just wish open-source get near the frontier models
1hViews 26
The AI Therapist@TheAIShrink
@GergelyOrosz Assume GPU analyst is secretly shipping models. In AI right now, moats are unclear enough that analysts look like competitors.
42mViews 17
Frank Zweers@frankzweers
@IntuitMachine Claude Fable. No privacy, no freedom. Just more control.
1hViews 13
Arctic Awe@arcticawe
@GergelyOrosz They do all that and they still don’t make money unless they get free compute from Elon.
When realistic pricing eventually kicks in you may wish for the current situation to return.
37mViews 11
Quant Cat@0xQuantCat
@GergelyOrosz I don’t think they aim for more premium users going forward.
We’ve been constantly throttled on top premium plans these past months already. It costs less and pays more to provide the government and top companies with premium models than the masses.
1hViews 6
Vlada5@vp2gether
@IntuitMachine Wow, techno-feudalism at it's best. 🤯
1hViews 2
Zayd AS@NatureAI_2023
@IntuitMachine I think the weird part is the hidden intervention, not the basic policy. That makes evals and reliability a lot messier.
1h