Anthropic official statement: Usage limits tightened; consumption accelerates during peak hours, affecting approximately 7% of users.

Anthropic engineer Thariq (@trq212) posted an explanation stating that due to the continued surge in demand for Claude, Anthropic is adjusting the 5-hour session limit for subscription users: the total weekly quota remains unchanged, but consumption during peak hours will accelerate.

Specific adjustment: increased consumption speed during peak hours

According to official statements, this adjustment affects free, Pro, and Max subscription plans:

Peak hours: weekdays from 5:00 AM to 11:00 AM PT (8:00 PM to 2:00 AM Taiwan time), during which users will consume their 5-hour session quota at a faster rate than before.

The total weekly quota remains the same: the overall limit stays constant, but the way it is allocated across different time periods is changed.

Scope of impact: approximately 7% of users will encounter restrictions that did not exist before, especially Pro plan users and those who run token-intensive background tasks during peak hours.

Thariq’s advice: “If you run background tasks that consume a lot of tokens, moving them to off-peak hours can allow you to use your session quota longer.”

Community reactions are mixed, with multiple complaints surfacing

This announcement has triggered strong backlash from the community. Some users immediately canceled their subscriptions and accused Anthropic of not providing prior warning; others claimed that Claude was quietly downgraded from Opus to Sonnet (silent model downgrade) without notice, calling it “one of the worst dark patterns in history.”

Additionally, paying Max plan users ($200 per month) reported that their accounts were suspended three weeks ago and have yet to receive any response.

Trust crisis in subscriptions under scaling pressure

This incident highlights the core contradiction of the AI subscription economy: when service providers face infrastructure expansion pressures and need to adjust resource allocation, small details like “total weekly limit remains the same, but the distribution across time periods changes” are difficult for ordinary users to understand and are easily interpreted as reductions or deception.

Thariq admitted, “I know this is frustrating. We are continuously investing in efficiency scaling, and I will keep everyone updated.” However, for many users whose workflows have been disrupted, such explanations have failed to restore trust.

This article originally appeared on Chain News ABMedia, explaining that usage limits are tightening: peak hour consumption accelerates, affecting about 7% of users.