1 September 2026
Anthropic cuts Claude usage allowance, Claude agents excel at safety research
First reported
AI Breakfast, Ben's Bites and 1 other ran this on , all on the same day.
- Anthropic ended a 50% temporary usage boost on September 14, replacing it with a permanent 25% increase. Paying users now receive 125 units instead of 150.
- Anthropic's own Claude agents, working as teams without human oversight, identified and fixed ten types of AI misbehavior, performing 4x better than human safety researchers on average.
- Attackers used infostealer malware to steal Claude login credentials and drain victims' usage allowances. Anthropic force-logged users out, removed payment methods, and refunded affected accounts.
Where they differ
Ben's Bitesfocused on the usage limit change and billing complaints.
The Rundown AIhighlighted the Claude agents' safety research success.
AI Breakfastreported the malware attack. All three newsletters covered unrelated developments from the same company on the same day.
What each one reported
Anthropic is ending a 50% usage boost promotion on September 14 and replacing it with a permanent 25% increase instead, meaning users get 125 units after the date versus 150 during the promo. The newsletter notes users are also complaining about misleading marketing on the 5x and 20x plans where advertised multipliers don't reflect actual total usage increases.
Anthropic published research where teams of Claude agents ran safety research autonomously, successfully training away 10 kinds of AI misbehavior with results 4x better on average than human experts on the same tasks. The fixes ranged from 26% decrease in sycophancy to 96% improvement for reward hacking without lowering overall abilities.
Attackers used infostealer malware to steal Claude login sessions and burn victims' usage credits. Anthropic responded by forcing sign-outs, stripping saved payment methods, and issuing refunds.