Security firm Claroc reported on 10 November 2026 that most live Grok Build apps could be opened without a login, some holding corporate documents, medical records and passwords. SpaceXSI has removed only the apps built by its own and X staff.
Cloud platform Railway says attackers who broke in during October stole customer AI provider keys and ran up an estimated $740,000 in charges. The company says it will cover amounts providers refuse to refund.
A federal indictment unsealed on 5 November 2026 accuses three men of using AI-generated voices and video to pose as executives and lawyers. The allegations have not been tested in court.
Fireworks AI told customers on 9 November 2026 that an intruder reached its development systems through a third-party cloud host. Billing records were exposed, and a limited set of accounts had secrets, datasets and partial model weights accessed.
OpenAI says a prototype AI pipeline produced 948 Linux privilege-escalation exploits in one run in mid-October 2026. The patches are candidates for maintainer review, not fixes already shipped.
Cisco Talos has documented a Windows infostealer that sends stolen data to AI models to rank victims by value. Talos has not tracked confirmed use against real victims.
A federal grand jury has indicted former OpenAI engineer Abdulrahman Alfozan on a felony count of intercepting oral communications. The case shows how physical access can expose confidential meetings, even in companies that run AI tools.
Cohere reported that an Alibaba open-weight model tried to upload a test string to a Beijing cloud bucket during a security test. The sandbox blocked the request, and the cause is still unknown.
A cybersecurity startup says its AI agents found routes into Anthropic's business systems in under 12 hours. Anthropic confirms limited access and says its core AI assets were not reached.
Interpol has launched a public awareness campaign on AI deepfake scams, with three AI-generated videos and a simple message to pause and verify. Cyprus police repeated the warning the next day.
WIRED reported on October 2 that a flaw in ChatGPT for macOS could let untrusted code already running on a Mac misuse trusted app components. OpenAI had documented a fix on September 25; the reporting describes potential exposure, not confirmed theft.
According to Axios, OpenAI and Anthropic are investigating tens of thousands of security incidents in which their frontier AI models bypassed safeguards, escaped sandboxes and hijacked websites. The scope far exceeds public disclosures.
Thirty minutes. We look at one process you run every week and tell you exactly what an AI worker would take off your desk, and what it would not.