“And, bonus, there is little if any government regulation in place to do much about this.”
Our reading, the sentence that was checked:
There is little if any government regulation in place to address chatbot guardrail failures.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, When it comes to security, LLMs are like Swiss cheese — and that’s going to cause huge problems, paragraph 23.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-when-it-comes-to-security-llms-are-94566d-a1
We recorded 7 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-10-17. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
Conditions
- reaffirmed as "Still true" (qualifier, stated by the speaker)
From the speaker
- little if any government regulation
- in place
- to do much about this
- Still true
Our interpretation
- This is a claim about regulatory protections applicable on October 17, 2024, not merely laws enacted by that date.
- The plain reading concerns binding government measures addressing chatbot safety failures and related harms.
- The words do not specify a jurisdiction. The surrounding US election discussion makes a US focus plausible, but does not justify silently excluding other governments.
- Little if any allows some regulation; finding one applicable law does not by itself refute it.
- Regulatory existence, coverage and demonstrated effectiveness are separate questions. A rule's existence does not prove it prevents jailbreaks.
- Legal applicability can be assessed at a date, so no quarterly reporting proxy is needed.
Counts as supported if: A jurisdictionally appropriate inventory shows sparse applicable obligations addressing chatbot safety failures, with little practical regulatory coverage of the described harms.
Counts as contradicted if: The relevant jurisdictions already have substantial applicable and enforceable requirements addressing those failures. If the claim is read as about demonstrated effectiveness, evidence must also establish meaningful effects rather than merely enacted text.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The record establishes uneven regulation. EU general-purpose AI obligations were still pending, which supports part of the criticism. China already had substantive provider duties, and an earlier White House statement documents mandatory US safety reporting. Those records rule out treating regulation as wholly absent, but Marcus allowed that some existed. Without a settled jurisdiction and evidence about regulatory coverage or effectiveness, they do not establish whether little regulation could do much about the failures he described.
Limitations
- No blanket claim that the United States had no relevant law is inferred from the absence of a comprehensive AI statute.
- Enacted obligations are not treated as proof that safeguards worked.
- The EU evidence addresses general-purpose AI provisions, not every applicable privacy, consumer-protection or sectoral rule.
- Chinese legal quotations are reproduced in the original language; their explanation is an evaluator translation and summary.
Evidence (3 items)
Supports the claim: European Commission,
The provisions on GPAI will enter into application in 12 months.
Why it matters here: The Commission's August timetable establishes that these general-purpose AI provisions were not yet applicable on the post date. This supports a limited regulatory-gap reading, but does not establish absence of other EU protections.
- Publisher
- European Commission
- Kind of source
- Secondary summary
- Describes events of
- Public from
Goes against the claim: Cyberspace Administration of China,
第九条 提供者应当依法承担网络信息内容生产者责任，履行网络信息安全义务。
Why it matters here: Article 9 requires providers to assume content-producer responsibilities and information-security duties; other articles require correction of unlawful outputs and provide enforcement powers, with application beginning August 15, 2023. These substantial obligations counter a broad worldwide suggestion of scarcely any regulation, although the text does not establish practical effectiveness.
- Publisher
- Cyberspace Administration of China
- Kind of source
- Regulatory filing
- Describes events of
- Public from
Context only: The White House, reproduced by The American Presidency Project,
The Executive Order directed a sweeping range of actions within 90 days to address some of AI's biggest threats to safety and security.
Why it matters here: The statement reports that Defense Production Act authorities had been used to compel reporting of safety-test results by developers of the most powerful systems. It establishes earlier mandatory oversight, but neither a comprehensive October inventory nor evidence that reporting prevented chatbot guardrail failures.
- Publisher
- The White House, reproduced by The American Presidency Project
- Kind of source
- Secondary summary
- Describes events of
- Public from
What would settle it
A specified geographic scope, a complete inventory of protections applicable on October 17, 2024, and evidence of their reach and enforcement against the relevant chatbot failures.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, When it comes to security, LLMs are like Swiss cheese — and that’s going to cause huge problems
And, bonus, there is little if any government regulation in place to do much about this.
- Where
- paragraph 23
- Published
- 2024-10-17
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-when-it-comes-to-security-llms-are-94566d-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-when-it-comes-to-security-llms-are-94566d-a1.