Anthropic's new model: too dangerous for the public

What's really behind Anthropic's withdrawal of its new model?

Last week, Anthropic launched a new version of its Claude model — unofficially nicknamed "Fable 5" by the community — and just days later pulled it from public access. The company cited safety reasons: the model was "too powerful" for the general public. However, sceptics see a commercial strategy rather than a genuine risk.

Until June 22, the model was available to subscribers. From that date, the plan was to move it to a pay-per-token system via the API, at a substantially higher cost. The sudden withdrawal caught many by surprise, and rumours are now circulating that access could be restored under different terms.

Safety or business?

The argument that the model was dangerous for non-technical users clashes with market logic: if a product is superior, the usual move is to sell it, not hide it. Several analysts suggest the decision may stem from performance problems or hallucinations — the model "hallucinated more than a zealot," according to a technical source — which the company preferred to iron out before exposing it to mass criticism.

Politics creeps into the debate

Some suggest Anthropic may have clashed with the Trump administration's interests — in mid-2026 — and that the withdrawal would be retaliation. Others see Elon Musk's hand at play, given Anthropic's infrastructure rental agreements with him. The fact is there is no official confirmation of any interference.

Meanwhile, the average user who was hoping to try the model is left empty-handed. And the industry gets yet another reminder that in artificial intelligence, what is free today may be paywalled tomorrow — or simply cease to exist.