Between January and August of this year, violent rhetoric aimed at AI and tech executives changed shape. It stopped announcing itself in a way traditional systems can catch.
Alethea reviewed tens of thousands of posts referencing AI executives across social platforms and news coverage during that period. Our model excludes things like obvious "trolls" and jokes, and instead looks at actionable threats of violence, meaning those with potential for online-to-offline action. For six months, the volume held steady at roughly one post a day containing violent rhetoric. Then, within a matter of weeks, it reached 255 posts a day — a 250-fold increase.
Most of that volume didn't use a single violent word or phrase. It showed up as an eponym, renamed tactic, a historical reference, or an image. This included everything from calls for a "Sam Altman special" (referencing the molotov cocktail thrown at his house in April 2026) to an actual replica guillotine towed to a city council hearing.
None of it would trip a keyword-based alert. That's the point. The language evolves after every incident and spreads into communities with no connection to AI. Increasingly, it extends past executives to the engineers, facility staff, and local officials tied to the same narrative.
Artemis for Security is built to catch that layer: coded language, memes that go beyond a bad joke, and coordinated patterns across accounts. Each signal is severity-scored and routed into the workflows your team already runs.
Our new report breaks down how this coded language works, why it's so difficult to monitor, where it's spreading, and what security and communications teams need to change about how they monitor for it.
Download the full report below, and reach out for a free risk assessment on the threats of violence targeting your organization.