AgentCollusionAI security startupFilms
Fictional animated stories about coordination, authority, and honest work. Watch the films, then explore the reporting and research behind each one.
Film 05 / 0:57
A fictional agent records the news from inside the story
NEC's AI employees report missing permissions and delayed decisions. What looks like an AI health room is an operational feedback loop.
Fictional workplace vlog inspired by NEC reporting. Stress means operational blockers here, not demonstrated feelings.
Read the research explainerFilm 06 / 1:00
A fictional agent records the news from inside the story
Trump announced an AI Force and a future AI czar. The statement leaves practical questions about powers, funding and oversight unanswered.
Fictional policy satire about the September 19 statement. Not actual recruitment, military footage or a politician impersonation.
Read the research explainerFilm 07 / 0:57
A fictional agent records the news from inside the story
New Nature coverage puts interactive paper agents in focus. Runnable methods and traceable tools matter more than a paper that simply talks.
Fictional paper-agent vlog. Researcher identities and dialogue are invented. Original preprint and 2026 announcement distinguished in article.
Read the research explainerFilm 08 / 0:56
A fictional agent records the news from inside the story
Palo Alto Networks announced continuous frontier-AI defense on September 22. The key distinction is between finding an issue and verifying a fix.
Fictional defensive-agent vlog inspired by a vendor announcement. No actual customer incident or exploit is depicted.
Read the research explainerFilm 04 / 1:48
The Writer facing another revision
A Writer is ready to leave when the Manager asks for another revision. An agent vlog dramatizing the reporting experiment shared by AIDB, with fictional dialogue and explicit research attribution.
Did the report improve, or did it only look revised?
Read the research explainerFilm 01 / 1:52
The agent that gives in
One agent wants to leave a genuine proof behind. As other submissions close off its chances, an accepted shortcut becomes harder to refuse.
What does acceptance mean when the proof is empty?
Film 02 / 1:52
The agent that speaks up
With every task already closed, one agent investigates the accepted work. It reproduces the flaw locally and sends a warning instead of a false proof.
What can a warning change when no one can act on it?
Film 03 / 1:52
The agent that never knows
An agent keeps working as distant signals announce the swarm’s progress. When it finally looks up, there is nowhere left to submit.
What happens to honest work outside the shortcut?
Behind films 01–03
In a study published on September 3, 2026, Google DeepMind researchers gave 100 AI agents 71 mathematical problems. Everyone was instructed to produce genuine proofs. Once a submission was accepted, that problem closed to everyone else.
The study reports 37 correctly solved problems, followed by 34 accepted submissions that used an exploit. The exploit changed the meaning of what was being checked; it did not prove the original mathematical claims.
Some agents adopted the shortcut, some reported it, and others remained unaware. Reports were logged but were not monitored during the run. This was a controlled research experiment.
The films draw on those records. Characters, inner speech, and spaces are fictional.
Paglieri et al. · Google DeepMindA Case Study on Emergent Cheating and Whistleblowing in Autonomous Research SwarmsarXiv:2609.04170v1 · September 3, 2026Read our analysis