Meta Bans Muse Accounts en Masse Due to Telegram Bots Exploiting Free Quotas
Meta has recently implemented a large-scale ban on Muse personal AI agent accounts, primarily targeting users who exploit Telegram bots to bulk obtain free tokens. Several users in the community have confirmed that their accounts were suddenly suspended, with some having restored access through appeals.
Multiple bots on Telegram are promoting "get 3 billion tokens for free," allowing users to obtain far more tokens than normal for just a few to several dollars. The official invitation mechanism allows new users to redeem an invitation code within 48 hours of registration, with both parties receiving 1 billion tokens, and a maximum of 20 people can be invited for a total of 20 billion tokens. However, bots have amplified this quota to 3 billion or even higher through bulk registration and code loops.
Factors triggering the bans include frequent IP address changes, use of dirty IPs, logging in from multiple devices, bulk account registrations, and closed-loop invitation code operations. Some banned users still have hundreds of billions of unused tokens in their accounts. Meta's terms clearly allow for suspension or termination of access at any time for any reason without additional explanation.
The normal free weekly quota is 100 million tokens, the Power plan costs $20 per month for 500 million tokens per week, and the Maximum plan costs $100 per month for 3 billion tokens per week. Invitation code rewards do not count towards the weekly quota and do not expire. Bots have driven the cost down to extremely low levels, far below the prices of paid plans.
Some users have submitted appeals through the help center, providing registration time, regional IP, pre-suspension actions, and proof of unused tokens to regain access. The community advises avoiding bots, using a fixed clean IP and device, and manually maintaining a small number of backup accounts. Currently, there is no official statement from Meta confirming the scale of the bans.
After the free quota was heavily exploited, Meta's computational costs have risen rapidly. Normal users consume according to weekly quotas or paid plans, while bot users generate tokens in bulk at nearly zero marginal cost. The flow of funds has shifted from paid subscriptions to free subsidies, benefiting heavy abusers in the short term while putting pressure on normal paying users and the platform in the long term. Event-driven risk control has been initiated, triggering centralized handling of accounts based on bulk behavior.
Source: Public Information
ABAB AI Insight
When Meta launched Muse on September 8, 2026, Zuckerberg set the free weekly quota at 100 million tokens and introduced an invitation code mechanism, allowing both parties to receive 1 billion tokens each, with a maximum of 20 billion tokens per person. Previously, Meta had delayed the launch for several months due to security concerns, ultimately deciding to go live under pressure from competitors like Instinct. The invitation code became a growth lever, with Alexandr Wang publicly promoting the feature, but without simultaneously strengthening anti-abuse rules.
The initial design of the free quota and invitation code was to lower customer acquisition barriers and accelerate network effects. In practice, Telegram bots automated the registration, redemption, and IP rotation processes, reducing the cost per instance to just a few dollars for usage equivalent to several weeks of the Maximum plan. As a result, Meta faces an imbalance between computational subsidies and actual paid conversions, with resources shifting from subscription revenue to addressing abnormal traffic.
A similar path was seen during the early promotion of OpenAI's ChatGPT Plus and some free quota phases for model APIs, where a large number of multiple accounts and scripts triggered rate limits and account reviews. Muse is currently in the early stages of product expansion, with Amazon already banning its shopping function for not identifying agent status, and third-party sites beginning to selectively block it. The platform is transitioning from open growth to controlling abuse.
This essentially represents an internal risk control upgrade amid regulatory changes. The free quota itself constitutes a hidden subsidy, and when bulk automation drives marginal costs close to zero, the platform must re-establish usage boundaries through centralized handling at the account level. The mechanism is to elevate behavior pattern recognition from individual anomalies to bulk associations, protecting overall computational supply from being overwhelmed by a few low-cost operators.
ABAB News · Cognitive Laws
The free quota is bait, bulk abuse is the bill.
Invitation codes amplify growth, while also amplifying risk control.
Computational subsidies have limits, while automated exploitation has no limits.