Grant De Swardt, an independent AI consultant, found his Claude account tokens mysteriously disappearing, despite no work being done. After a partial refund and account suspension, Anthropic confirmed a compromised session key, possibly due to malware. De Swardt's experience echoes those of others, who faced similar unauthorized usage. The lack of itemized usage tracking and the difficulty in securing accounts have led many to switch to alternative AI services.

De Swardt's frustration is palpable. His sole proprietor business, reliant on AI, saw its operations disrupted. Despite warnings from Anthropic, he found no evidence of his own computer being compromised. His subscription was ultimately cancelled in favor of Cursor, citing better token management and affordability.

The episode highlights the increasing risk of AI misuse and the need for more robust security measures. Anthropic's response, while commendable, falls short in providing clear guidance on how users can identify and prevent such incidents.

The incident underscores the broader challenge of safeguarding personal data in an era where AI is integral to daily operations. For now, users are left wondering how to protect their accounts and ensure their AI tools operate as intended.