WorkOS, Northflank and Trigger.dev disclose AI-linked breaches

Three technology vendors — WorkOS, Northflank and Trigger.dev — have disclosed hacks linked to the theft of AI access tokens, according to a SecurityWeek report published on 4 November 2026.

SecurityWeek said hackers have been harvesting AI access tokens and have been using AI-assisted offensive tools to automate attacks. The three disclosures were reported together as part of what the outlet described as a mounting set of AI-linked breaches.

Access tokens are credentials that let software call a service on a user's behalf, without re-entering a password. In an agent context, those tokens are what allow an automated system to keep working while nobody is watching. That makes them attractive to steal: a token lifted from one system can often be replayed against another until it is revoked.

The pattern described by SecurityWeek has two halves. The first is the harvesting of credentials tied to AI services. The second is the use of AI-assisted offensive tooling to automate the attacks themselves, which lowers the manual effort involved in moving from one compromised account to the next.

Detail remains limited. The report does not set out how many customers of each vendor were affected, what data was reached, or when the intrusions began. The incidents were reported as confirmed and disclosed by the companies themselves, with the disclosures surfacing on 4 November 2026. No attribution to a particular group has been reported in the material available, and no motive has been established.

For organisations running agents that hold long-lived tokens, the practical questions raised by the disclosures are familiar ones: where tokens are stored, how long they live, what scopes they carry, and how quickly they can be revoked across every system that trusts them. Where a breach involves a supplier rather than the organisation itself, the answers often depend on the supplier's own rotation and notification process.

This report is based on a single published account. The Agentic Times will update the story if the vendors publish fuller incident write-ups, including timelines, the number of tokens exposed and any guidance on rotation. Customers of the three companies should watch their vendors' own security advisories for instructions rather than relying on secondhand summaries, since revocation steps and affected scopes are specific to each platform.