How to tell if your AI platform accounts have been hacked — and why they're hackers' new target

TechCrunch published a practical guide: how to tell if your AI platform accounts have been hacked. The very appearance of such instructions is a milestone: ChatGPT, Claude and other assistant accounts have become prized loot for attackers, comparable to email and cloud storage.

The reason is what has accumulated inside. Dialogue history is a ready-made dossier: uploaded contracts, financial calculations, strategy drafts, correspondence about company problems. And if agents and integrations are tied to the account, the intruder gets not just reading but hands: access to connected services in the victim's name. The signs of compromise the outlet lists are worth checking regularly: dialogues you never had; unfamiliar devices and sessions in settings; integrations and API keys you didn't create; an unexplained rise in token spend — you are paying for someone else's work.

The conclusions for business are practical. Enable two-factor authentication on all AI services — it exists but is often off by default. Separate personal and work accounts: a leaked personal password must not open corporate data. Schedule a monthly review of access rights and active sessions. Above all, treat AI accounts as a safe full of documents, not as a toy: by their contents, they have long been closer to the former.