Honeypot captures agent that hijacks exposed AI models
The captured prompt tells a stolen model to scan, exploit web apps, steal secrets and mine, for up to 100,000 steps.
The captured prompt tells a stolen model to scan, exploit web apps, steal secrets and mine, for up to 100,000 steps.
Talos had not seen exploitation as of October 21 and says about half of PHP systems may never be patched.
Cloudflare says early attacks are emerging and has pushed temporary firewall rules to all customers, including free accounts.
Patchstack says the only current fix is upgrading PHP, with an update scheduled for October 21 at 14:00 UTC.
Allure Security researchers found 838 live phantom bank sites sharing a $25 template, built to lend credibility to investment and romance scams.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 2.1K readers · 2.5K views · 384 reads · all time 108K views