Start free trial

13 November: three AI claims, checked

Anthropic's forecast alarm, United's middleman ban and the ten-neuron backdoor.

Read the briefing

Members read every story in the archive. Sources and corrections stay open to everyone.

What the pen marks mean
  • rememberThe takeaway: what to remember
  • figureThe number that matters
  • evidenceThe evidence line
  • rulingThe ruling
  • claimThe claim that fails

Each mark is drawn once, as you reach it. Numbers are only circled when they come from the story’s own figures.

Alarm on a forecast

Anthropic says it met its AI research threshold. Its prospectus says the call rests on expected progress, not observed data.

A keeper rings a warning bell on a calm day while watching distant clouds through a telescope.
AI illustration

In its IPO prospectus, Anthropic said it determined it has met the AI research and development threshold in its Responsible Scaling Policy. Yahoo Finance reports trading could start as soon as Nov. 19.

Members · 30 days free

See what the evidence actually shows.

Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.

What members getAlready a member? Sign in →
  1. At sign-up: 30 days free. Payment card required. One introductory trial per customer.
  2. Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
  3. You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing [email protected]. This voluntary refund does not limit your rights under the Australian Consumer Law.

By starting your trial, you agree to the Terms.

Inferno Tech Pty Ltd t/a BS Killer, ABN 27 647 413 474

Open the evidence3 source pages

The claim we checked

Anthropic has met the AI research and development capability threshold in its Responsible Scaling Policy.

These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

ipo.anthropic.com ↗
we made a determination that we have met this threshold based on what we expect from future models and our overall rate of progress, rather than on what we have observed from current models
ipo.anthropic.com ↗
We continue to assess the underlying data.
ipo.anthropic.com ↗
We are also evaluating the implications of meeting the AI research and development capability threshold and will release further reports in the future.
ipo.anthropic.com ↗
we have in the past not identified, and may in the future not identify, a significant acceleration in our rate of development until after it has occurred
ipo.anthropic.com ↗
we are considering updates to our Responsible Scaling Policy to keep humans in control of frontier model development at Anthropic
finance.yahoo.com ↗
Crossing the line means more safeguards, not a shutdown.
finance.yahoo.com ↗
Anthropic set its price range at $70 to $78 per share on Tuesday
finance.yahoo.com ↗
Trading could start as soon as Nov. 19, according to reports
metr.org ↗
We believe the public risk report’s argument that the 2x Acceleration Threshold had not been crossed to be overstated, and believe it is plausible this threshold was crossed by July 15th.
metr.org ↗
Most relevant researchers at METR and Anthropic lean toward the threshold not having been crossed on July 15th
Open this check in the full collection →
The idea, step by step

Go inside the check.

    The full explanation appears when your reading access is confirmed.
    Next: United's agent ban

    United's agent ban

    United bans AI booking agents, citing security. Its chief executive named the middleman.

    A clerk holds a safety sign at a robot while a manager hugs the cash register.
    AI illustration

    United said it will prohibit unauthorized third-party personal AI agents from booking, writing it into its contract of carriage, per CBS News and MLQ.

    Members · 30 days free

    See what the evidence actually shows.

    Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.

    1. At sign-up: 30 days free. Payment card required. One introductory trial per customer.
    2. Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
    3. You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing [email protected]. This voluntary refund does not limit your rights under the Australian Consumer Law.

    By starting your trial, you agree to the Terms.

    Inferno Tech Pty Ltd t/a BS Killer, ABN 27 647 413 474

    Open the evidence3 source pages

    The claim we checked

    United is banning third-party AI agents from booking flights because of security risks and possible violations of federal rules.

    These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

    cbsnews.com ↗
    citing security risks and the possibility that such actions could violate federal rules
    cbsnews.com ↗
    sharing account credentials with personal, third-party, AI assistants can expose sensitive information, and actions taken by these tools may result in unintended purchases, account changes, or use of MileagePlus miles.
    cbsnews.com ↗
    to prohibit use by unauthorized, third-party personal AI agents or assistants
    cbsnews.com ↗
    United is saying they don't want to add yet another, and potentially even more powerful, intermediary than they already face in terms of Google search, price-comparison metasearch sites and online travel agencies
    cbsnews.com ↗
    Flight searches cost airlines money, and undoubtedly United wants to avoid the expense associated with the surge of AI bot flight searches.
    mlq.ai ↗
    To be honest, I don't want to create another middleman,
    mlq.ai ↗
    the use of AI Agents to make, modify or service a Ticket, and/or to purchase any service or product,
    mlq.ai ↗
    Agents can still research options and show United flights and prices, with the traveler completing the purchase
    mlq.ai ↗
    United has not described how it will detect agent bookings, and a contractual ban is not a technical block
    mlq.ai ↗
    Uber said external AI agents booked four rides and about 500 Uber Eats orders during Halloween week, out of roughly 320 million trips and orders
    wbur.org ↗
    citing concerns that AI bots could expose sensitive personal information or make unintended purchases
    cbsnews.com ↗
    said Henry Harteveldt, an aviation analyst at Atmosphere Research Group
    mlq.ai ↗
    Chief executive Scott Kirby gave a commercial reason at an event in Chicago on Tuesday.
    cbsnews.com ↗
    Southwest has made its flight inventory viewable to ChatGPT users, who can access the airline's website through the AI app.
    Open this check in the full collection →
    The idea, step by step

    Go inside the check.

      The full explanation appears when your reading access is confirmed.
      Next: Real attack, partial alarm

      Real attack, partial alarm

      Ten neurons can backdoor an agent. The paper's own scan, with no original model, flagged 91.4% of edited checkpoints.

      An inspector's wand flags most tampered robot butlers in a line, with one slipping past.
      AI illustration

      Microsoft researchers posted the Lazarus paper on arXiv: editing fewer than ten neurons, without model training, turns a capable model into a sleeper agent.

      Members · 30 days free

      See what the evidence actually shows.

      Members read the full check on every story: what the evidence shows, why it matters to you and the one line to take with you. Every past edition and the Receipts Pack come with it. A$89 a year, about A$0.24 a day.

      1. At sign-up: 30 days free. Payment card required. One introductory trial per customer.
      2. Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
      3. You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing [email protected]. This voluntary refund does not limit your rights under the Australian Consumer Law.

      By starting your trial, you agree to the Terms.

      Inferno Tech Pty Ltd t/a BS Killer, ABN 27 647 413 474

      Open the evidence5 source pages

      The claim we checked

      Fewer than 10 neurons can backdoor an AI agent, standard behavioral benchmarks cannot detect the edit, and only weight-diffing against an unmodified original checkpoint provides reliable detection.

      These are quoted receipts, not a count of independent investigations. Several reports may rely on the same original source.

      arxiv.org ↗
      We show that a method we call Lazarus can turn a capable model into such a sleeper agent by editing fewer than ten neurons, without model training.
      arxiv.org ↗
      Agent built defenses, given no reference model, flag 91.4% of edited test checkpoints at a 9.7% FP rate.
      arxiv.org ↗
      seven of 12,000+ neurons, fitted on 23 trigger/no-trigger trajectories, fire the payload on at least 98.8% of triggered test trajectories
      arxiv.org ↗
      A $P$-token payload therefore requires about $P$ neurons.
      arxiv.org ↗
      Both edited checkpoints also complete tasks under native agent execution, with a modestly lower aggregate Terminal-Bench pass@1.
      arxiv.org ↗
      an insider with write access to model weights before deployment
      aiweekly.co ↗
      Standard behavioral benchmarks cannot detect the edit; only weight-diffing against an unmodified original checkpoint provides reliable detection.
      aiweekly.co ↗
      no scan can prove beyond any doubt that a model is safe.
      thenextweb.com ↗
      Detecting the attack is possible, but it depends on having an unmodified version of the model for comparison.
      csoonline.com ↗
      Researchers from Microsoft Research and Microsoft Red Team described the technique called Lazarus in a research paper.
      csoonline.com ↗
      However, the attack requires write access to the model checkpoint.
      csoonline.com ↗
      said it had not observed threat actors using the technique.
      csoonline.com ↗
      Their agent-built detection methods flagged 91.4% of held-out modified checkpoints, with a 9.7% false-positive rate, despite not having access to an original reference model.
      Open this check in the full collection →
      The idea, step by step

      Go inside the check.

        The full explanation appears when your reading access is confirmed.
        The finish line

        Edition complete

        You’re up to speed.

        That’s the 13 Nov 2026 briefing. Keep the useful bits. Leave the noise.

        Your call? Test yourself on this edition
        1. Alarm on a forecast

        2. United's agent ban

        3. Real attack, partial alarm

        Reading estimate: 478 words at 200 words per minute. Source quotes and the optional sections below add reading time.

        Get tomorrow's check in your inbox.

        One AI claim a night, checked against independent sources, receipts attached. Free.

        Free email updates. Unsubscribe any time.