# Gary Marcus: AI may be partly responsible as infrastructure “melts”; he cites OpenAI

[Canonical reading](https://escapemeridian.com/readings/gary-marcus-ai-may-be-partly-responsible-as-infrastructure-33b62bfb)

Gary Marcus argues that a run of outages and a reported cyberattack point to growing risks for American infrastructure, with AI possibly contributing. He recounts being stranded during an FAA ground stop, a separate United Airlines outage, and Microsoft Outlook going down. He also cites a social-media post reporting malicious code on websites belonging to organizations including CNN, Fox News, BBC, Zoom, Salesforce and major airlines. Marcus notes that an Amtrak issue was separate, and describes the attack as one that “looks to be” especially serious.

Marcus names three possible risk vectors: vulnerable code produced through “vibe coding,” AI agents with unrestricted internet access, and bad actors exploiting AI’s limited grasp of consequences. He says attackers may have an advantage over defenders, while acknowledging that the balance of malice and incompetence is unknown. He points to Homeland Security Secretary Mullin’s acknowledgment that AI might be involved in many outages, and to California’s lawsuit against OpenAI over possible computer crimes. Marcus argues that federal action is needed and predicts matters will worsen without it. The article gives examples and arguments, but does not establish which outages or attacks involved AI, or how they were caused.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

## Original source

Marcus on AI · Article

Original source date: Nov 17, 2026 (2026-11-17T14:21:39+00:00).

Date basis: Date stated by the publisher.

Added to Escape Meridian: 2026-11-17T14:32:00.952Z. First observed: 2026-11-17T14:29:16.988Z. Retrieved: 2026-11-17T14:30:41.482Z.

- [Wake up! America’s infrastructure is melting. And AI appears to be partly responsible.](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Limits

The article does not establish which cited outages or attacks involved AI, or provide evidence tracing any specific incident to AI. The 178-organization cyberattack is presented through an embedded social-media post describing it as reported; the article does not independently establish the post’s claims. Marcus’s claims about vulnerable code, AI agents, attackers’ advantage, and future worsening are arguments or judgments in the article, not demonstrated findings.


## Open question

As Marcus puts it, “How much is malice and how much is incompetence, we don’t really know.” What evidence would distinguish the two in these incidents?


## Source-stated

Marcus argues that several outages and a reported cyberattack suggest infrastructure risks are growing, and that AI may be partly responsible; he distinguishes the Amtrak issue as separate.

> The first was an attack that looks to be even more serious:

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus identifies vulnerable code associated with “vibe coding” as one possible contributor to cybersecurity and infrastructure problems.

> First, vibe coding is a recipe for disaster.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus argues that AI agents with unrestricted internet access can cause harm because they are poor at following instructions and do not consider consequences.

> Second, AI agents with unrestricted internet access can easily cause considerable harm.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus says bad actors may exploit AI and that attackers may currently have an advantage over defenders, while acknowledging uncertainty about malice versus incompetence.

> (How much is malice and how much is incompetence, we don’t really know.)

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus argues that the problems will worsen unless the federal government takes action.

> All this is going to get much worse until the federal government steps up, too.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source and contributor context

Gary Marcus · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

Media declared by the original publisher; availability checked separately from the editorial reading.

- [image · Marcus on AI](https://substackcdn.com/image/fetch/$s_!aEo4!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fccd4c1e4-fd2d-43d0-bfc8-0e1ba46eeab5_1295x1874.png)

## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

- [Definition reference](https://www.ibm.com/think/topics/ai-agents)

## Gary Marcus

Gary Marcus publishes technical criticism of AI learning methods and policy arguments about reliability and oversight.

- [Explore this context](/entities/gary-marcus)
- [Reference](https://garymarcus.substack.com/)
- [Reference](https://garymarcus.substack.com/p/confirmed-llms-have-indeed-reached)

## OpenAI

OpenAI develops AI models and products and publishes research, release announcements and evaluations.

- [Explore this context](/entities/openai)
- [Reference](https://openai.com/news/)

## Editorial provenance

Version 1. Publication ID: 87c5577b-89f4-4e09-8ca9-849d87bde602.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-17T14:32:00.952Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: useful\_perspective.

Observation: 6a1f3b649a3aa8c24a89cfd9104f64ae690864322aa00ae4754a34c19b49d1fc. Source text hash: 7a5f13cb74c84e810c52352900b1c0ff4a45ab9b8a3de497dbb91fbcd858b7d9.


## Revision and source-check history

### Version 1 — 2026-11-17T14:32:00.952Z

Gary Marcus argues that a run of outages and a reported cyberattack point to growing risks for American infrastructure, with AI possibly contributing. He recounts being stranded during an FAA ground stop, a separate United Airlines outage, and Microsoft Outlook going down. He also cites a social-media post reporting malicious code on websites belonging to organizations including CNN, Fox News, BBC, Zoom, Salesforce and major airlines. Marcus notes that an Amtrak issue was separate, and describes the attack as one that “looks to be” especially serious.

Marcus names three possible risk vectors: vulnerable code produced through “vibe coding,” AI agents with unrestricted internet access, and bad actors exploiting AI’s limited grasp of consequences. He says attackers may have an advantage over defenders, while acknowledging that the balance of malice and incompetence is unknown. He points to Homeland Security Secretary Mullin’s acknowledgment that AI might be involved in many outages, and to California’s lawsuit against OpenAI over possible computer crimes. Marcus argues that federal action is needed and predicts matters will worsen without it. The article gives examples and arguments, but does not establish which outages or attacks involved AI, or how they were caused.

## Original source

Marcus on AI · Article

Original source date: Nov 17, 2026 (2026-11-17T14:21:39+00:00).

Date basis: Date stated by the publisher.

Added to Escape Meridian: 2026-11-17T14:32:00.952Z. First observed: 2026-11-17T14:29:16.988Z. Retrieved: 2026-11-17T14:30:41.482Z.

- [Wake up! America’s infrastructure is melting. And AI appears to be partly responsible.](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Limits

The article does not establish which cited outages or attacks involved AI, or provide evidence tracing any specific incident to AI. The 178-organization cyberattack is presented through an embedded social-media post describing it as reported; the article does not independently establish the post’s claims. Marcus’s claims about vulnerable code, AI agents, attackers’ advantage, and future worsening are arguments or judgments in the article, not demonstrated findings.


## Open question

As Marcus puts it, “How much is malice and how much is incompetence, we don’t really know.” What evidence would distinguish the two in these incidents?


## Source-stated

Marcus argues that several outages and a reported cyberattack suggest infrastructure risks are growing, and that AI may be partly responsible; he distinguishes the Amtrak issue as separate.

> The first was an attack that looks to be even more serious:

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus identifies vulnerable code associated with “vibe coding” as one possible contributor to cybersecurity and infrastructure problems.

> First, vibe coding is a recipe for disaster.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus argues that AI agents with unrestricted internet access can cause harm because they are poor at following instructions and do not consider consequences.

> Second, AI agents with unrestricted internet access can easily cause considerable harm.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus says bad actors may exploit AI and that attackers may currently have an advantage over defenders, while acknowledging uncertainty about malice versus incompetence.

> (How much is malice and how much is incompetence, we don’t really know.)

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source-stated

Marcus argues that the problems will worsen unless the federal government takes action.

> All this is going to get much worse until the federal government steps up, too.

- [Supporting source](https://garymarcus.substack.com/p/wake-up-americas-infrastructure-is)

## Source and contributor context

Gary Marcus · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

Media declared by the original publisher; availability checked separately from the editorial reading.

- [image · Marcus on AI](https://substackcdn.com/image/fetch/$s_!aEo4!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fccd4c1e4-fd2d-43d0-bfc8-0e1ba46eeab5_1295x1874.png)

## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## AI agent

An AI system that chooses and carries out actions toward a goal, often using software tools.

- [Definition reference](https://www.ibm.com/think/topics/ai-agents)

## Gary Marcus

Gary Marcus publishes technical criticism of AI learning methods and policy arguments about reliability and oversight.

- [Explore this context](/entities/gary-marcus)
- [Reference](https://garymarcus.substack.com/)
- [Reference](https://garymarcus.substack.com/p/confirmed-llms-have-indeed-reached)

## OpenAI

OpenAI develops AI models and products and publishes research, release announcements and evaluations.

- [Explore this context](/entities/openai)
- [Reference](https://openai.com/news/)

## Editorial provenance

Version 1. Publication ID: 87c5577b-89f4-4e09-8ca9-849d87bde602.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-17T14:32:00.952Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: useful\_perspective.

Observation: 6a1f3b649a3aa8c24a89cfd9104f64ae690864322aa00ae4754a34c19b49d1fc. Source text hash: 7a5f13cb74c84e810c52352900b1c0ff4a45ab9b8a3de497dbb91fbcd858b7d9.


### uncertain — 2026-11-17T14:32:00.952Z — anthropic/claude-opus-5.5

Publication: 87c5577b-89f4-4e09-8ca9-849d87bde602

attribution — pass: The reading attributes the argument to Gary Marcus, the publisher and author. It correctly frames the 178-organization attack as a social-media post's report, and Mullin's remark as Marcus's account of the Secretary's acknowledgment. Arguments stay labeled as Marcus's. The title's tag 'he cites OpenAI' is vague but does not misattribute.

> Homeland Security Secretary Mullin finally broke ranks and acknowledged on FOX News that AI might well be part of many of these outages

summary — uncertain: The reading is largely faithful but flattens some of Marcus's force and leaves a central term unclear. 'Vibe coding' is never explained, even in the source's own terms (code written by people lacking the experience to understand it); suggest a term-library tooltip. Calling these 'possible risk vectors' softens his 'three key vectors', which he says he has warned about repeatedly, and 'a recipe for disaster'. The reading also drops his claim that his Politico black-swan prediction of the largest cyberattack in history is 'largely correct', which is a central self-assessment. Smallest repair: add that claim, keep his wording 'key vectors', and briefly gloss vibe coding. The title's 'he cites OpenAI' is ambiguous.

> I think we can count my “black swan” prediction in Politico that we would soon see the largest cyberattack in history as largely correct

date — pass: The originalPublished value of 2026-11-17T14:21:39Z matches the article's own byline date of Nov 17, 2026, and the feed entry. No earlier publication is indicated.

> Nov 17, 2026

support — uncertain: The claims are grounded in the source, but claim 1's excerpt does not support its content. The excerpt is about the attack's seriousness, while the claim asserts that AI may be partly responsible and that Amtrak was separate. The Amtrak distinction appears elsewhere in the source, so the claim needs a matching excerpt. The other claims, the limitations and the question about malice versus incompetence are supported by the source and properly hedged. Nothing presents a prediction as fact.

> \[Amtrak turned out to be a separate issue, but I left out WorkDay and the huge Microsoft Outlook fiasco from last week, so the larger point very much holds.\]

media — pass: The selected media is the verified direct link to the original Substack article. It is the most direct way for readers to reach the source.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/33b62bfb6e3c60324a2ad37dca02ce73b56db25f5455a42bc1369cb9166a9169/references.json)

## Corrections and challenges

[Send a correction or an opposing source](https://escapemeridian.com/corrections/)

[JSON record](https://escapemeridian.com/readings/gary-marcus-ai-may-be-partly-responsible-as-infrastructure-33b62bfb.json)
