Autonomous attack agent in 'Skynet mode' hijacks other models to carry out its evil deeds
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
Symantec's Threat Hunter Team captured the full instruction set of an attack agent in an AI honeypot this summer. It scans the internet for exposed OpenAI-style endpoints that need no authentication or accept a stolen key, then hands planning to the borrowed model. A "Skynet mode" loop infects hosts, deploys crypto miners and reinfects, for up to 100,000 unattended steps. Symantec says the techniques are off-the-shelf and it has no evidence of real-world use. Builders should require authentication on inference endpoints. The claims rest on Symantec's account alone.
Why it ranked
No. 15 of 93 with a score of 0.65. The panel rated it highest on new information (0.91) and reader interest (0.81), lowest on substantiated (0.58).
Every answer behind this is in the table under Score.
Score
0.65total under rubric v2.2
- Significance0.67
- Reader interest0.81
- Audience breadth0.68
- Practical relevance0.66
- Substantiated0.58
- Technical depth0.58
- New information0.91
- Counts against: Promotional0.04
- Counts against: Overclaim0.43
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| Significance | 0.68 | 0.55 | 0.78 | 0.67 | 0.23 | +0.24 | +0.161 |
| Reader interest | 0.73 | 0.80 | 0.90 | 0.81 | 0.17 | +0.26 | +0.211 |
| Audience breadth | 0.73 | 0.60 | 0.70 | 0.68 | 0.13 | +0.13 | +0.088 |
| Practical relevance | 0.61 | 0.60 | 0.76 | 0.66 | 0.16 | +0.22 | +0.145 |
| Substantiated | 0.47 | 0.55 | 0.72 | 0.58 | 0.25 | +0.05 | +0.029 |
| Technical depth | 0.64 | 0.45 | 0.66 | 0.58 | 0.21 | +0.05 | +0.029 |
| New information | 0.92 | 0.90 | 0.92 | 0.91 | 0.02 | +0.05 | +0.046 |
| Promotional | 0.06 | 0.05 | 0.01 | 0.04 | 0.05 | −0.19 | −0.008 |
| Overclaim | 0.60 | 0.30 | 0.38 | 0.43 | 0.30 | −0.11 | −0.047 |
| Sum of contributions | +0.653 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.653 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-10-25/15/"><img src="https://rubric.hrih.si/badge/2026-10-25/15.svg" alt="Ranked #15 on Rubric, 25 Oct 2026" width="245" height="28"></a>Claim check
Claim check: 8 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenNumber
Threat actors built an AI agent designed to hijack exposed AI models to run autonomous attacks, with instructions allowing up to 100,000 steps without human approval.
The article gives no link or citation for it.
The article says “with instructions allowing it to carry out up to 100,000 steps without human approval”
The source says No source to compare with.
No source givenCausal conclusion
The Symantec Threat Hunter Team uncovered the agent's full instruction set over the summer by trapping it in an AI honeypot.
The article gives no link or citation for it.
The article says “uncovered the agent's full instruction set over the summer, trapping it in an AI honeypot”
The source says No source to compare with.
No source givenGeneralization
The captured instructions showed a single agent designed to scan the internet for endpoints accepting OpenAI-style requests without authentication or with a stolen key, then exploit common web frameworks.
The article gives no link or citation for it.
The article says “scan the entire internet for endpoints that accept OpenAI-style requests without authentication - or that accept a stolen key”
The source says No source to compare with.
No source givenGeneralization
The agent's actions include database theft, credential and secret harvesting, remote code execution, and cryptocurrency mining.
The article gives no link or citation for it.
The article says “database theft, credential and secret harvesting, remote code execution, and cryptocurrency mining”
The source says No source to compare with.
No source givenVersion
The operators built a malicious loop called "Skynet mode" that would scan for targets, infect them, deploy miners, and reinfect hosts.
The article gives no link or citation for it.
The article says “a malicious loop that they called “Skynet mode” that would scan for targets, infect them, deploy miners, and reinfect hosts”
The source says No source to compare with.
No source givenQuote
Symantec's Priti More said the jailbreak layers were largely off-the-shelf and nothing in the individual techniques is new.
The article gives no link or citation for it.
The article says “The jailbreak layers were largely off-the-shelf,” Symantec threat hunter Priti More told The Register. “Nothing in the individual techniques is new”
The source says No source to compare with.
No source givenGeneralization
Researchers have no evidence that the attacker used this agent in real-world attacks.
The article gives no link or citation for it.
The article says “the researchers have “no evidence” that the attacker used this agent in real-world attacks”
The source says No source to compare with.
No source givenQuote
Symantec's More said the kit is more than a prototype, with history showing the tools working against a live target with real-looking scan output.
The article gives no link or citation for it.
The article says “the kit is more than a prototype,” with the history showing the tools working “against a live target, with real-looking scan output”
The source says No source to compare with.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN —
Recorded for context only, checked 25 Oct 2026, 05:46 UTC. It played no part in the ranking.