AI
“Former OpenAI and Anthropic researcher Jacob Coxon claims China already has spies inside both OpenAI and Anthropic, stating he is 90% certain despite providing no evidence." (Instagram, @therundownai, published 2026-10-09)”
Plain restatementJacob Coxon, who previously worked as a researcher at OpenAI and at Anthropic, publicly stated that he is approximately 90 percent confident that Chinese intelligence operatives are already employed inside OpenAI and Anthropic, and he did not present supporting evidence. Secondary elements: he warned China could steal US lab advances and could hack the labs; China technology analyst Paul Triolo called the allegation irresponsible; Coxon left Anthropic in September over AI safety concerns.
Distortion code this site does not recognise yet: capability_extrapolation. Not collectible until the field guide has an entry.
This post accurately reports what was said. Jacob Coxon, who worked as a researcher at OpenAI and then Anthropic before resigning in September 2026 over AI safety concerns, said on the PBD Podcast in early October that China already has spies inside OpenAI and Anthropic, and when asked how certain he was he answered ninety percent. Multiple named news outlets, including the South China Morning Post, reported the remarks and noted that he gave no evidence and named no one. A China technology analyst, Paul Triolo, publicly criticised the allegation as irresponsible, which the post also includes. The important separate point is that the allegation itself remains completely unverified: there is no public evidence, no company confirmation, no named individual, and no legal case supporting it, and neither OpenAI nor Anthropic appears to have responded publicly. Do not confuse this with OpenAI's unrelated dismissal of three employees on 1 October 2026, which the company attributed to mishandling sensitive information shared with an outside AI evaluation group, not to foreign espionage. In short, the post correctly describes a claim that one person made, and that claim has not been shown to be true.
[drifted from the evidence:] Former OpenAI and Anthropic [drifted from the evidence:] researcher Jacob Coxon claims China already [drifted from the evidence:] has spies inside [drifted from the evidence:] both OpenAI and Anthropic, [drifted from the evidence:] stating he [drifted from the evidence:] is 90% certain despite providing no evidence." [drifted from the evidence:] (Instagram, @therundownai, published 2026-10-09)
[added by the neutral restatement:] Jacob Coxon, who previously worked as a researcher at OpenAI and [added by the neutral restatement:] at Anthropic, [added by the neutral restatement:] publicly stated that he is approximately 90 percent confident that Chinese intelligence operatives are already [added by the neutral restatement:] employed inside OpenAI and Anthropic, [added by the neutral restatement:] and he [added by the neutral restatement:] did not present supporting evidence. [added by the neutral restatement:] Secondary elements: he warned China could steal US lab advances and could hack the labs; China technology analyst Paul Triolo called the allegation irresponsible; Coxon left Anthropic in September over AI safety concerns.
Red-tinted words in the claim drifted from the evidence. Green-tinted words are what a neutral restatement needs.
The trace / claim to source
- Coxon did make the statement. Multiple named outlets, including SCMP and Business Insider as relayed by a wire aggregator, report him saying on the PBD Podcast that China already has spies at OpenAI and Anthropic.
- The ninety percent figure is accurately quoted. Reporting records the host asking how certain he was and Coxon answering with ninety percent and the words almost certain.
- The post's "providing no evidence" qualifier is accurate and matches SCMP's own wording that he made the allegation without providing evidence. He named no individual and produced no document.
- Coxon's described background is accurate. He worked at OpenAI and then Anthropic, by his own account three years of pretraining research across both.
- He resigned from Anthropic in September 2026 over AI safety concerns, announced publicly on 8 September 2026, a fact confirmed by NPR, Axios and TIME.
- The hacking element is fairly restated. Reporting quotes him saying China has the ability to hack at any time and could take US progress.
- The Triolo criticism is supported by reporting, which quotes him on X describing Coxon's remarks as "really irresponsible."
- SCMP did publish an article on this, so the post's source credit points to a real piece.
- Capability extrapolation, in minor form and only in the image text rather than the caption: the on-image headline reads as a flat assertion that a former insider says China has spies in both labs, without the ninety percent and no-evidence qualifiers that appear in the caption. A viewer who sees only the graphic takes away an insider revelation rather than an insider's personal probability estimate. The caption corrects this.
- The post does not state what the underlying allegation's evidentiary status is beyond "no evidence." A reasonable reader may not register that no company, agency, court filing, or named investigator has corroborated any part of the infiltration claim, and that Coxon's statement is an inference about likelihood rather than a report of something he observed.
- Whether Chinese intelligence operatives are in fact inside OpenAI or Anthropic is unverified. No public evidence, company confirmation, prosecution, or regulator finding supporting the allegation was found.
- Whether Coxon has any non-public basis for the estimate is unknown. Reporting indicates he offered reasoning about weak vetting rather than specific knowledge, but the full podcast was not reviewed, so the complete basis he gave cannot be characterised beyond what outlets quoted.
- Neither OpenAI nor Anthropic appears to have publicly responded to this specific allegation. Their silence is neither confirmation nor denial.
- The precise division of Coxon's tenure between the two companies is not established by the strongest sources. His own public statement describes three years across both; a more granular split appears only in social-media accounts, which are weak sourcing.
Multiple named outlets report the same underlying event: Coxon appeared on the PBD Podcast hosted by Patrick Bet-David in early October 2026 and said that China already has spies at OpenAI and Anthropic. Asked by the host how certain he was, he gave a figure of ninety percent and said it was almost certain. He added that China has the ability to hack the labs at any time and could take US progress and catch up, and that there is no serious background check process for this yet. SCMP reports the ninety percent figure and states explicitly that he gave the allegation without providing evidence. Other outlets covering the same remarks report that he named no individuals and offered no documentation, and that there has been no public confirmation of the allegation. Coxon's prior record is well documented. He resigned from Anthropic on 8 September 2026 with a public post saying he had spent three years doing pretraining research at OpenAI and Anthropic and that neither company was acting responsibly, and he told Axios he left before his equity vested. He has since appeared on broadcast programmes and testified before the New York City Council on 5 October 2026 about AI risk. On the criticism element, reporting states that China technology analyst Paul Triolo wrote on X that Coxon's remarks were "really irresponsible." Separately, and not the same event, OpenAI said on 1 October 2026 that it had parted ways with three employees over mishandling sensitive company information, including sharing it with an outside AI evaluation organisation. The dismissed employees dispute the characterisation and say they were punished for safety advocacy. No reporting links those dismissals to Chinese espionage.
Complete reasoning
The reply is formatted for pasting into the thread where the claim is circulating.
Compact share page: ai.trueseeker.com/s/b7be56e3faa3/MX-ihMNAxMoH5VZaIs9jQicEQUi
Ask this case
Answers come only from the case file above; nothing is added.
Did Jacob Coxon actually say China has spies inside OpenAI and Anthropic?
Yes. Multiple news outlets report that Coxon said this on the PBD Podcast in early October 2026, and when asked how certain he was, he said ninety percent.
Did Coxon provide any evidence for his claim?
No. Reporting, including from the South China Morning Post, states he made the allegation without providing evidence, named no individuals, and produced no documentation.
Who is Jacob Coxon and why does his opinion carry weight?
He worked as a researcher at OpenAI and then Anthropic, doing about three years of pretraining research across both companies, and resigned from Anthropic in September 2026 citing AI safety concerns.
Is it actually true that China has spies in these AI labs?
That is not established. There is no public evidence, company confirmation, prosecution, or regulator finding supporting the claim, so it remains unverified.
Did OpenAI or Anthropic respond to the allegation?
The investigation found no public response from either company to this specific claim. Their silence does not confirm or deny it.