Analysis
Microsoft says Onyx Sleet produced dozens of tailored lures in minutes. In the same October 29 assessment, Jade Sleet launched campaigns through three front companies in one week.
Those examples come from Microsoft Threat Intelligence, which describes growing AI use by North Korean threat actors: assistance with target research, deceptive identities, software tooling and repetitive operational work. The company explicitly assesses that the operations remain human-directed.
That's where the cases split.
The finding concerns acceleration of established activity. It is not evidence that an autonomous system has taken over an entire attack.
Microsoft distinguishes levels of adoption among named groups. The Jade Sleet and Onyx Sleet cases are the company's intelligence findings, not a public census of every operator or every campaign.
This is one vendor's assessment. The observations and the adoption categories are independently unverified, and they sit inside Microsoft's visibility and attribution limits.
Not established: the effectiveness of any particular defensive product, or how much risk any one organization faces.
I would not treat a convincing job application, business introduction or technical exchange as proof of effort. The checkpoint is the request that extends trust: share a repository, run a tool, change a payment destination, or grant an account access.
Recruiting, engineering and security may each see a plausible fragment and miss the sequence. The report strengthens the case for following that handoff across the teams.
Who verifies the sender on a channel that did not come with the message? Who sees the whole path when an ordinary exchange becomes a privileged request? Who can refuse the repository, the tool, the payment change or the account?
Source: Tracking AI adoption in North Korean threat actor tradecraft · 2026-10-29
Why it matters
Microsoft's account of Jade Sleet and Onyx Sleet weakens polished prose as a sign of legitimacy, and leaves the real check on the trust handoff.

