“Already, the advent of general purpose agents with broad internet access has led to tens of thousands of hacking incidents”
Our reading, the sentence that was checked:
General purpose agents with broad internet access have already caused tens of thousands of hacking incidents.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Coming soon: New York City’s hearing on AI risks, paragraph 9.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-coming-soon-new-york-citys-hearing-27ff37-a2
We recorded 12 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 2026-10-05. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Causal claim
- 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
- agents with broad internet access (qualifier, stated by the speaker)
From the speaker
- Already
- general purpose agents with broad internet access
- has led to
- tens of thousands of hacking incidents
Our interpretation
- The claim concerns cumulative incidents occurring by October 5, 2026.
- Read 'tens of thousands' as at least 20,000 distinct incidents.
- A hacking incident involves unauthorized access or an actual attempt against a computer system. It need not cause damage. Messages, evaluation runs, policy violations and individual requests are not automatically separate hacking incidents.
- The agents must have materially performed or enabled the hacking, rather than merely being present.
Counts as supported if: Credible incident records establish at least 20,000 distinct hacking incidents attributable to internet-connected general-purpose agents by publication.
Counts as contradicted if: A sufficiently comprehensive accounting establishes a total below that threshold. A smaller documented sample alone does not contradict the claim.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Does the evidence support the cause?
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 September 28, 2026 report describes a broad aggregate that includes adversarial tests and safeguard bypasses; it does not establish 20,000 distinct hacking incidents by October 5, 2026 [e1]. Anthropic's September 9, 2026 investigation documents four incidents of unauthorized access to real third-party systems, but its search does not establish an industry-wide total or upper bound [e2]. These sources therefore neither establish the claimed threshold nor show that the total falls below it [e1, e2].
Limitations
- The aggregate is available through secondary reporting and does not separate distinct hacking attempts from simulated tests and other behavior [e1].
- Anthropic's investigation covers its own model activity and cannot establish a total for all general-purpose agents [e2].
Evidence (2 items)
Context only: International Business Times UK,
a broad category of behaviour
Why it matters here: The article describes an aggregate that includes unsuccessful attempts and adversarial tests. It does not supply a count of distinct real-world hacking incidents, including attempts, under the contract.
- Publisher
- International Business Times UK
- Kind of source
- Secondary summary
- Describes events of
- Public from
Context only: Anthropic,
four incidents
Why it matters here: The company documents real unauthorized access during evaluations. Its investigation establishes a limited set of incidents, not an industry-wide total or upper bound.
- Publisher
- Anthropic
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
A dated, deduplicated accounting establishing at least 20,000 qualifying incidents by October 5, 2026, or a sufficiently comprehensive accounting establishing a total below that threshold. It must distinguish actual hacking attempts and unauthorized access from simulated tests and other behavior, and establish that internet-connected general-purpose agents materially performed or enabled each incident.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Coming soon: New York City’s hearing on AI risks
Already, the advent of general purpose agents with broad internet access has led to tens of thousands of hacking incidents
- Where
- paragraph 9
- Published
- 2026-10-05
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-coming-soon-new-york-citys-hearing-27ff37-a2: 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-coming-soon-new-york-citys-hearing-27ff37-a2.