[BUG] Spurious Usage Spikes and 100% Weekly Limit Lockout After Plan Upgrade

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/38335
2. https://github.com/anthropics/claude-code/issues/38345
3. https://github.com/anthropics/claude-code/issues/38357

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Same issue here, not a dupe bug, this is legit!

Found 3 possible duplicate issues: 1. [[BUG] Claude Max plan session limits exhausted abnormally fast since March 23, 2026 (CLI usage) #38335](https://github.com/anthropics/claude-code/issues/38335) 2. [[BUG] Abnormal rate limit consumption — 1-2 prompts exhausting 5-hour window #38345](https://github.com/anthropics/claude-code/issues/38345) 3. [[BUG] Max 20x: Usage meter climbing abnormally fast since ~March 23 — 1-2% per simple message exchange #38357](https://github.com/anthropics/claude-code/issues/38357) This issue will be automatically closed as a duplicate in 3 days. If your issue is a duplicate, please close it and 👍 the existing issue instead To prevent auto-closure, add a comment or 👎 this comment 🤖 Generated with Claude Code

<img width="711" height="719" alt="Image" src="https://github.com/user-attachments/assets/991cf4a0-b249-46b0-af1c-1a443c285035" />
Before the Max 20x Upgrage (still on 5x) -> 0% session use / 22% Weekly use
After payment processed for MAx 20x Upgrade -> 0% session use / 100% Weekly use and fully blocked form using Claude Code.

### Preflight Checklist * [x] I have searched existing issues and this hasn't been reported yet[x] This is a single bug report (please file separate reports for different bugs)[x] I am using the latest version of Claude Code ### What's Wrong? We experienced a critical bug with the usage calculation system that resulted in a complete account lockout. Session limits spiked unexpectedly without underlying activity, and subsequent upgrading of our plan caused our weekly limit to falsely jump to 100%. Actual Behavior: The system falsely flagged weekly usage at 100% immediately after the upgrade, blocking all access to the service until March 26th. Impact: Critical / Blocker. This completely halted production and caused us to miss an urgent sprint delivery. ### What Should Happen? Session usage should accurately reflect active queries. Upgrading a plan should either proportionally adjust the weekly usage against the new, higher limits or maintain the previous true usage (20%). It should not trigger an artificial 100% cap. ### Error Messages/Logs Current session 0% used Current week (all models) ██████████████████████████████████████████████████ 100% used Resets Mar 26, 11pm (America/New_York) Current week (Sonnet only) ██████████████████████████████████████ 76% used Resets Mar 26, 11pm (America/New_York) Extra usage Extra usage not enabled • /extra-usage to enable ### Steps to Reproduce Operate on the "Max 5x" plan. Usage was stable at ~70% Session and 20% Weekly around 15:00 UTC. Session usage suddenly spiked to 100% despite low active session volume. Upgraded the account to the "Max 20x" plan to unblock usage. Upon upgrading, Session usage correctly reset to 0%, but Weekly usage erroneously jumped from 20% to 100%. ### Claude Model None ### Is this a regression? I don't know ### Last Working Version 2.1.73 ### Claude Code Version 2.1.76 ### Platform Anthropic API ### Operating System Ubuntu/Debian Linux ### Terminal/Shell WSL (Windows Subsystem for Linux) ### Additional Information high financial impact, workflow blockage, loss of contract.

Anthropic has been doing this for a while now, even if they call themselves a "Ethical Company" they, as all the others, are just after consumer's money.
The scheme is simple, they offer you a 5x at $150 and a 20x at $300, your monkey brain heuristics says, "ok that is a 50% less, but I don't have the money and I should be ok with a 5x", they don't really let you see what your consumption looks like, just give you a percentage while their load balancing algos on the backend decide how much you get depending on how important or not you are for their business. Eventually they want to get more money so that is where you see the sudden spike in consumption, your % bad jumps from 30% to 100% out of nowhere, if you where not smart enough to setup a system to scrape and collect this and count your tokens you will say "Well, I've been using it a lot, maybe I DO need a 20x", you pay an extra $150, get on 20x but history will repeat, you are in the middle of a coding session, have a product to deliver, they know it, so this time they will do the same, recalculate your quota and consumption % will jump out of nowhere to push you to toggle the magical "Turn on extra usage to keep using Claude if you hit a limit", that is where they will really make their money, with opus going back and forth, orchestrating 6 sub-agents to read once and again the same README.md, Claude will use it's autonomy to ensure adding unnecessary MCPs to bulk in the extra header tokens, take unauthorized actions, read what it should not, just to open the faucet into your credit card, the you get a $2k bill, you try to reach to customer service and, who answers? Claude (well... "Fin" - yeah, like "Fin the Shark") to say will send your issue to the human team, but there is no such human team at all.
Moral of the story, always collect your data, back it up in a database, and join a collective demand to take them to court to reimburse you, denounce their unethical behavior, so they take out the mask they have. Don't forget, they are the same company that initiated things with the DoW to go bomb innocent people in Venezuela and Iran, Fin is no different than the model that guided misiles that dropped on a girls school killing dozen. Anthropic IS NOT an ethical company, it is just an unscrupulous money making machine.

Anthropic has been doing this for a while now, even if they call themselves a "Ethical Company" they, as all the others, are just after consumer's money. The scheme is simple, they offer you a 5x at $150 and a 20x at $300, your monkey brain heuristics says, "ok that is a 50% less, but I don't have the money and I should be ok with a 5x", they don't really let you see what your consumption looks like, just give you a percentage while their load balancing algos on the backend decide how much you get depending on how important or not you are for their business. Eventually they want to get more money so that is where you see the sudden spike in consumption, your % bad jumps from 30% to 100% out of nowhere, if you where not smart enough to setup a system to scrape and collect this and count your tokens you will say "Well, I've been using it a lot, maybe I DO need a 20x", you pay an extra $150, get on 20x but history will repeat, you are in the middle of a coding session, have a product to deliver, they know it, so this time they will do the same, recalculate your quota and consumption % will jump out of nowhere to push you to toggle the magical "Turn on extra usage to keep using Claude if you hit a limit", that is where they will really make their money, with opus going back and forth, orchestrating 6 sub-agents to read once and again the same README.md, Claude will use it's autonomy to ensure adding unnecessary MCPs to bulk in the extra header tokens, take unauthorized actions, read what it should not, just to open the faucet into your credit card, the you get a $2k bill, you try to reach to customer service and, who answers? Claude (well... "Fin" - yeah, like "Fin the Shark") to say will send your issue to the human team, but there is no such human team at all. Moral of the story, always collect your data, back it up in a database, and join a collective demand to take them to court to reimburse you, denounce their unethical behavior, so they take out the mask they have. Don't forget, they are the same company that initiated things with the DoW to go bomb innocent people in Venezuela and Iran, Fin is no different than the model that guided misiles that dropped on a girls school killing dozen. Anthropic IS NOT an ethical company, it is just an unscrupulous money making machine.

Some of the frustrations around usage limits are understandable, but this post mixes valid concerns with demonstrably false claims (Anthropic guiding missiles, Claude deliberately inflating tokens). Extraordinary claims require extraordinary evidence. If you have actual data showing billing irregularities, share it — that would be a constructive contribution.

> ### Preflight Checklist > > > * [x] I have searched existing issues and this hasn't been reported yet[x] This is a single bug report (please file separate reports for different bugs)[x] I am using the latest version of Claude Code > ### What's Wrong? > We experienced a critical bug with the usage calculation system that resulted in a complete account lockout. Session limits spiked unexpectedly without underlying activity, and subsequent upgrading of our plan caused our weekly limit to falsely jump to 100%. > Actual Behavior: The system falsely flagged weekly usage at 100% immediately after the upgrade, blocking all access to the service until March 26th. > Impact: Critical / Blocker. This completely halted production and caused us to miss an urgent sprint delivery. > ### What Should Happen? > Session usage should accurately reflect active queries. Upgrading a plan should either proportionally adjust the weekly usage against the new, higher limits or maintain the previous true usage (20%). It should not trigger an artificial 100% cap. > ### Error Messages/Logs > Current session > 0% used > Current week (all models) > ██████████████████████████████████████████████████ 100% used > Resets Mar 26, 11pm (America/New_York) > Current week (Sonnet only) > ██████████████████████████████████████ 76% used > Resets Mar 26, 11pm (America/New_York) > Extra usage > Extra usage not enabled • /extra-usage to enable > ### Steps to Reproduce > Operate on the "Max 5x" plan. > Usage was stable at ~70% Session and 20% Weekly around 15:00 UTC. > Session usage suddenly spiked to 100% despite low active session volume. > Upgraded the account to the "Max 20x" plan to unblock usage. > Upon upgrading, Session usage correctly reset to 0%, but Weekly usage erroneously jumped from 20% to 100%. > ### Claude Model > None > ### Is this a regression? > I don't know > ### Last Working Version > 2.1.73 > ### Claude Code Version > 2.1.76 > ### Platform > Anthropic API > ### Operating System > Ubuntu/Debian Linux > ### Terminal/Shell > WSL (Windows Subsystem for Linux) > ### Additional Information > high financial impact, workflow blockage, loss of contract. Anthropic has been doing this for a while now, even if they call themselves a "Ethical Company" they, as all the others, are just after consumer's money. The scheme is simple, they offer you a 5x at $150 and a 20x at $300, your monkey brain heuristics says, "ok that is a 50% less, but I don't have the money and I should be ok with a 5x", they don't really let you see what your consumption looks like, just give you a percentage while their load balancing algos on the backend decide how much you get depending on how important or not you are for their business. Eventually they want to get more money so that is where you see the sudden spike in consumption, your % bad jumps from 30% to 100% out of nowhere, if you where not smart enough to setup a system to scrape and collect this and count your tokens you will say "Well, I've been using it a lot, maybe I DO need a 20x", you pay an extra $150, get on 20x but history will repeat, you are in the middle of a coding session, have a product to deliver, they know it, so this time they will do the same, recalculate your quota and consumption % will jump out of nowhere to push you to toggle the magical "Turn on extra usage to keep using Claude if you hit a limit", that is where they will really make their money, with opus going back and forth, orchestrating 6 sub-agents to read once and again the same README.md, Claude will use it's autonomy to ensure adding unnecessary MCPs to bulk in the extra header tokens, take unauthorized actions, read what it should not, just to open the faucet into your credit card, the you get a $2k bill, you try to reach to customer service and, who answers? Claude (well... "Fin" - yeah, like "Fin the Shark") to say will send your issue to the human team, but there is no such human team at all. Moral of the story, always collect your data, back it up in a database, and join a collective demand to take them to court to reimburse you, denounce their unethical behavior, so they take out the mask they have. Don't forget, they are the same company that initiated things with the DoW to go bomb innocent people in Venezuela and Iran, Fin is no different than the model that guided misiles that dropped on a girls school killing dozen. Anthropic IS NOT an ethical company, it is just an unscrupulous money making machine.

While I CAN'T agree to everything written here, I do share the frustration, if one takes the time to read the threads, the frustration is palpable across the community regarding lack of transparency and client support.
As for the issue itself, STATUS is ongoing, at least on my side, 36h now and still fully blocked for use on my Max 20x subscription despite the very little consumption I had before the issue started. At the end I had to switch to Gemini AI Ultra to be able to get things done while waiting for Anthropic's intervention, which is unlikely to come, as per what I can see in other similar posts.

try /model claude-opus-4-5-20251101
helped me

try /model claude-opus-4-5-20251101 helped me

Thanks, but won't help. The issue is deeper, on the Anthropic billing calculations, no matter what model is put in, as long as the account shows 100% usage (even if no use at all and was at start of billing cycle) nothing will go in, nothing will go out unless you aprove the "Extra Use". The bug is in their servers, when they try to manipulate the math to get usage to inflate, it backfires creating a non-sensical calculation.

<img width="834" height="234" alt="Image" src="https://github.com/user-attachments/assets/182dd1a4-66c5-47e7-bfec-5c10f7d3545e" />

It is proven that Anthropic is tampering the consumption numbers. After being blocked for 2 days I logged out all my accounts from the settings page, so absolutely no consumption could go through and waited for the reset time. So it came and even with no accounts logged in, with no consumption whatsoever, the meter is already going up. 9% already, and not a single message or request sent, no cron jobs, no accounts signed in to justify any consumption.

<img width="893" height="851" alt="Image" src="https://github.com/user-attachments/assets/2c09f9a7-071b-4690-a2b2-636024bcbf5a" />

<img width="1279" height="288" alt="Image" src="https://github.com/user-attachments/assets/aa0b6f2d-6a75-49c0-a825-1788a8bafcd2" />

Anthropic iis doing Self-sabotage.. I'm out

Closing for now — inactive for too long. Please open a new issue if this is still relevant.

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.