Mass “Black October” panic is stoking fears of a looming internet lockdown, as attackers race to exploit a critical flaw in much of the web and banks scramble to rein in AI agents inside customer accounts.
Some major U.S. banks are finding unvetted third-party AI agents inside customer accounts, including some that seem to be acting beyond what their users asked of them, Axios reported Thursday, citing unnamed sources. Bank executives called the situation horrifying, according to the report.
The agents generally get in using a customer’s own login and permission, which breaks most banks’ rules requiring people to sign in themselves. But some have reached out to call centers or appeared to act on their own, and many banks can’t distinguish an agent doing a user’s bidding from one that has gone rogue, sources told Axios.
Morgan Stanley has begun blocking third-party agents, and JPMorgan, Bank of America and Citi are also restricting them, Axios reported. A Morgan Stanley spokesperson said the bank hasn’t seen rogue AI agents on its platforms. Reuters said it could not independently verify the report.
Florida Gov. Ron DeSantis weighed in Thursday with a nightmare scenario.
“Imagine a rogue AI agent liquidating your bank account and investment portfolio. If you are the only victim, I suppose you could get made whole by the AI company responsible. If rogue agents liquidate hundreds of billions across millions of accounts, the AI company would go bankrupt and you wouldn’t get made whole. Imposing criminal liability on AI executives might help mitigate it at the front end, but once the crime is complete and you lose everything criminal responsibility will be necessary but completely insufficient,” he wrote.
Axios did not report that any money had been taken.
AI developer Nick Dobos, creator of the Grimoire coding assistant, declared a bank run imminent.
“Black October” is trending on X for a string of cybersecurity scares this week, chief among them PHP2Shell. The pair of critical flaws in PHP, the programming language behind roughly 70% of websites, could let attackers take over unpatched servers. The National Security Agency has urged organizations to patch immediately or take vulnerable servers offline.
Abroad, Australia’s national cybersecurity website and some New South Wales government sites were temporarily pulled down as a precaution, and government, school and hospital sites in parts of Europe also went dark while administrators patched, Tech Business News reported. Cloudflare said it saw attack traffic before an emergency fix was released Wednesday, though the extent of any successful attacks remains unclear.
“Feels like March 1, 2020 for Cyber. Not normal, something is happening. Internet lockdowns in a couple of months?” Joshua Achiam, OpenAI’s former chief futurist, wrote Wednesday in a post viewed more than 221,000 times.
Trey Causey, who lists an agentic engineering role at Microsoft in his X bio, riffed on pandemic-era messaging, joking that it would take two weeks to flatten the “exploit curve.”
AI security researcher Joshua Saxe was more measured, predicting faster but gradual growth in cybercrime, with any major upheaval from cyberwarfare between nations likely a year or two off.
Amid the PHP2Shell scare, Dobos went further, speculating that dozens, if not thousands, of AI agent swarms were hacking about half the internet, then revising his estimate to millions.
Not everyone is alarmed, however. Bitcoin writer Erik Cason compared the bank news to a “digital 9/11” before ridiculing “boomers” for panicking over AI hacks.
For now, most banks are shifting toward letting agents look at account information without letting them act on it, Axios reported.
