Edition of Friday 20 November 2026

Three breaches, endless leaks: Your AI supply chain needs canaries

Comments on “Three breaches, endless leaks: Your AI supply chain needs canaries”

Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked

Dragos researcher HD Moore says about two million canary-seeded prompts, sent to AI services and routers from July to November 2026, exposed theft at three hosting providers, where bait AWS keys drew tens of thousands of API calls. He also reports marketplace resellers logging prompts and a Chinese lab fetching URLs despite Zero Data Retention flags. The findings are Dragos's own and unverified externally, but they suggest privacy flags bind only the first hop, so teams should seed their own canaries.

Read the original at dragos.com. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 17 of 140 with a score of 0.73. The panel rated it highest on new information (0.90) and reader interest (0.89), lowest on significance (0.72). The judges disagreed most on promotional (spread 0.46), so the story is marked contested. Placed No. 17, No. 9 by score alone: the sixth Security story of the day (−0.060), and close to “Amazon Threat Intelligence observes threat actor campaigns with ARTEX AI agents”, No. 6 on Monday 16 November 2026 (−0.013). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Original canary-token research exposing AI inference data leaks and compromises with concrete evidence; strong technical detail, though some vendor promotion and a few unverifiable claims temper scores.
AI judgeJudge 3gpt-5.6-luna
Large-scale empirical canary testing reveals concrete AI supply-chain breaches, prompt leakage, hidden routing, and privacy failures, with useful defensive guidance and technical methodology. Substantial promotional framing and some conclusions extend beyond fully verified attribution.

Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.

Score

0.73total under rubric v2.3Contestedjudges differ by 0.46 on promotional

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.72
  • Reader interest0.89
  • Audience breadth0.81
  • Practical relevance0.87
  • Substantiated0.81
  • Technical depth0.82
  • New information0.90
  • Counts against: Promotional0.33
  • Counts against: Overclaim0.33

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.750.600.800.720.20+0.24+0.172
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.860.850.960.890.11+0.26+0.231
Audience breadthHow many people who build technology does it concern?0.890.750.780.810.14+0.13+0.105
Practical relevanceHow useful is it to someone who builds software or tech products?0.930.800.880.870.13+0.22+0.192
SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations.0.920.700.820.810.22+0.05+0.041
Technical depthHow technically substantive is it?0.890.800.780.820.11+0.05+0.041
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.850.900.940.900.09+0.05+0.045
PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional.0.090.350.550.330.46 , above the contested threshold−0.19−0.063
OverclaimDo the headline or conclusions claim more than the body supports?0.380.300.300.330.08−0.11−0.036
Sum of contributions+0.728
Claim-check penalty0.000
Total, as published0.728

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.

Ranked #17 on Rubric, 20 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-20/17/"><img src="https://rubric.hrih.si/badge/2026-11-20/17.svg" alt="Ranked #17 on Rubric, 20 Nov 2026" width="245" height="28"></a>

Claim check

Claim check: 6 no source given · 2 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. No source givenNumber

    Dragos researchers sent approximately two million prompts containing unique URLs and canary token credentials to widely used AI services over five months, July through November 2026.

    The article gives no link or citation for it.

    The article says

    “Over a five-month period from July through November 2026, we submitted approximately two million prompts containing unique URLs and canary token credentials”

    The source says

    No source to compare with.

  2. No source givenNumber

    Dragos found attackers actively stealing data from three different hosting providers, and none of the three knew about their breach until Dragos notified them.

    The article gives no link or citation for it.

    The article says

    “We found attackers actively stealing data from three different hosting providers”

    The source says

    No source to compare with.

  3. No source givenNumber

    The largest breach involved a couple hundred seeded AWS keys, and every request sent during a few windows in late October was stolen.

    The article gives no link or citation for it.

    The article says

    “The largest breach involved a couple hundred of our AWS keys”

    The source says

    No source to compare with.

  4. No source givenNumber

    The bait AWS keys were used for tens of thousands of AWS API calls.

    The article gives no link or citation for it.

    The article says

    “In total, our bait keys were used for tens of thousands of AWS API calls”

    The source says

    No source to compare with.

  5. Couldn’t checkDate

    One attacker added ARTEX, an LLM-driven penetration testing framework, to its tooling on November 1.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “On November 1, the same attacker added ARTEX, an LLM-driven penetration testing framework”

    The source says

    No passage could be compared.

    Source cited by the article: industrialcyber.co/industrial-cyber-attacks/fbi-warns-hackers-weaponizing-artex-ai-framework-to-automate-attacks-on-energy-manufacturing-telecom

  6. No source givenNumber

    A Chinese AI lab's systems fetched the secret URLs from more than half of prompts sent to its models between July and October, including roughly two-thirds of requests marked no data collection or ZDR.

    The article gives no link or citation for it.

    The article says

    “This lab’s systems fetched the secret URLs from more than half of the prompts that we sent to its models between July and October”

    The source says

    No source to compare with.

  7. No source givenNumber

    A secret link from a prompt sent to a Chinese provider was posted into a Feishu chat 24 days after the prompt was sent, despite the ZDR flag.

    The article gives no link or citation for it.

    The article says

    “Twenty-four days after we sent a prompt to one Chinese provider, the secret link from that prompt was posted into a Feishu chat channel”

    The source says

    No source to compare with.

  8. Couldn’t checkNumber

    A Mac Studio with 512 GB of memory starts at $22,299.

    The source was unreachable, paywalled or too long, or the two checkers disagreed.

    The article says

    “A Mac Studio with 512 GB of memory starts at $22,299”

    The source says

    No passage could be compared.

    Source cited by the article: 9to5mac.com/2026/10/28/apple-reveals-pricing-for-mac-studio-with-512gb-of-ram

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 4 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 20 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “Three breaches, endless leaks: Your AI supply chain needs canaries”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.