Skip to main content
NOPE
Low reported severity Criminal Charges

S.D.N.Y.: Judge Rakoff rules that a fraud defendant's written exchanges with Claude about his defence, seized by the FBI, are protected by neither attorney-client privilege nor work product

Bradley Heppner, a corporate executive charged in the Southern District of New York with securities fraud, wire fraud and related offences, used Anthropic's Claude in 2025, after receiving a grand jury subpoena, to prepare reports outlining his defence strategy. His counsel had not directed him to do so. FBI agents seized about thirty-one documents memorialising those exchanges when they searched his home in connection with his arrest in November 2025. Heppner claimed privilege over them. On 10 February 2026 Judge Jed S. Rakoff granted from the bench the Government's motion for a ruling that the documents were not protected from Government inspection by either the attorney-client privilege or the work product doctrine, and a memorandum filed 17 February 2026 gave the reasons: Claude is not an attorney, and the communications were not confidential under Anthropic's privacy policy. The court described the question as one of first impression nationwide. Heppner was convicted in May 2026 of securities fraud, wire fraud and other charges; no inspected source says whether the Claude documents were used at trial.

AI system
Claude
Anthropic
Occurred
10 Feb 2026
Reported
17 February 2026
Event location
United States
What the AI did
Communicated with the person
Reported harm
Legal Harm
Whose AI use
Their own AI use
Setting
Justice
Evidence
AI involvement supported · Causal attribution established · 4 sources
5 claims: 4 documented, 1 reported. 3 open questions
People reported harmed
1 person

AI system as recorded: Claude (Anthropic), a publicly available generative AI platform the defendant used on his own to prepare defence-strategy reports

What Happened

The indictment, returned on 28 October 2025, charged Heppner with securities fraud, wire fraud, conspiracy, making false statements to auditors and falsifying corporate records in connection with GWG Holdings. According to the memorandum, FBI agents executing a search warrant at his home on his arrest seized documents and devices, and his counsel told the Government that about thirty-one of the seized documents memorialised communications with Claude. Counsel said the exchanges took place in 2025, after Heppner received a grand jury subpoena and knew he was the target, and that without any suggestion from counsel he had 'prepared reports that outlined defense strategy'. Counsel conceded that it 'did not direct [Heppner] to run Claude searches.' Heppner asserted privilege on the grounds that he had put information learned from counsel into Claude, created the documents to speak with counsel, and later shared them with counsel. The court held that the documents lacked at least two elements of attorney-client privilege: Claude is not an attorney, and the exchanges were not confidential because Anthropic's privacy policy lets it collect inputs and outputs and disclose them to third parties. It held work-product protection unavailable because the documents were not prepared at the behest of counsel. i24NEWS, describing the ruling, says the defendant opened 31 separate chats and entered information he had received from his attorneys.

Reported harm

The defendant's self-prepared defence-strategy documents, written with Claude, lost the protection he claimed and were ruled open to Government inspection in his criminal case (court memorandum). Whether they were used at his trial is not reported.

Outcome

Resolved

The court granted the Government's motion from the bench on 10 February 2026 and filed a memorandum of reasons on 17 February 2026 (ECF No. 27, 25 Cr. 503 (JSR)). Under a Privilege Protocol Stipulation the Government and Heppner's counsel had agreed that the Government would segregate the documents and not inspect them pending that ruling. The Washington Post reports that Heppner was convicted in May 2026 of securities fraud and wire fraud, among other charges; his attorneys did not respond to its request for comment.

What remains unknown

  • Whether the Government used the Claude documents at trial or in plea or sentencing proceedings.
  • What the documents contained beyond counsel's description as defence-strategy reports.
  • Whether any appeal challenges the privilege ruling.

What the evidence supports

AI involvement: supported. The court memorandum records counsel's representation that the seized documents memorialise Heppner's communications with Claude and that he used it to prepare defence-strategy reports.

5 claims: 4 documented, 1 reported. What the statuses mean

Documented On 10 February 2026 the court granted from the bench the Government's motion for a ruling that Heppner's written exchanges with a generative AI platform were not protected from Government inspection by either attorney-client privilege or the work product doctrine; a memorandum of reasons followed on 17 February 2026.

Causal attribution. Established by the court's own memorandum.

  • storage.courtlistener.com(opens in new tab) supports · English
    [Page 1] 'At a pretrial conference in this matter held on February 10, 2026, the Court, after hearing the arguments of counsel, granted from the bench the Government's motion for a ruling that certain written exchanges that defendant Bradley Heppner had with a generative artificial intelligence ("AI") platform were not protected from Government inspection by either the attorney-client privilege or the work product doctrine.'
  • storage.courtlistener.com(opens in new tab) supports · English
    [Page 12] 'Because Heppner's use of Claude fails to satisfy either of these rules, the AI Documents do not merit the protections Heppner has claimed.'
  • yahoo.com(opens in new tab) supports · English
    'But the judge said that protections for conversations between clients and their attorneys did not apply: Claude is not a lawyer, and the executive's attorneys did not ask him to talk to the chatbot.'
Documented FBI agents seized about thirty-one documents memorialising Heppner's communications with Claude when they executed a search warrant at his home in connection with his arrest in November 2025 (the memorandum gives 4 November on page 3 and the day after the 4 November unsealing on page 2).

Causal attribution. Court record of counsel's representation; not a claim about AI behaviour.

  • storage.courtlistener.com(opens in new tab) supports · English
    [Page 3] 'In connection with Heppner's arrest on November 4, 2025, agents with the Federal Bureau of Investigation executed a search warrant at Heppner's home and seized numerous documents and electronic devices.'; 'Heppner's counsel later represented to the Government that among the seized materials were approximately thirty-one documents that memorialize communications that Heppner had with the generative AI platform "Claude," which is operated by the private company Anthropic.'
  • i24news.tv(opens in new tab) context · English
    'A criminal suspect consulted the Claude artificial intelligence tool, opening 31 separate chats and entering sensitive details about his case, including information he had received from his attorneys. The FBI later raided his home and seized printouts of the conversations.'
Documented According to his counsel, Heppner used Claude in 2025, after receiving a grand jury subpoena, to prepare reports outlining defence strategy, without any suggestion from counsel.

Causal attribution. The memorandum records counsel's own statements.

  • storage.courtlistener.com(opens in new tab) supports · English
    [Page 3] 'Without any suggestion from counsel that he do so, Heppner "prepared reports that outlined defense strategy, that outlined what he might argue with respect to the facts and the law that we anticipated that the government might be charging."'; [Page 4] 'Heppner's counsel conceded, however, that counsel "did not direct [Heppner] to run Claude searches."'
Documented The court held that the documents were not privileged because Claude is not an attorney and the exchanges were not confidential.

Causal attribution. Established by the court's reasoning.

  • storage.courtlistener.com(opens in new tab) supports · English
    [Page 5] 'Because Claude is not an attorney, see ECF No. 23-6, that alone disposes of Heppner's claim of privilege.'; [Page 6] 'Second, the communications memorialized in the AI Documents were not confidential.'
Reported Heppner was convicted in May 2026 of securities fraud and wire fraud, among other charges.

Causal attribution. No inspected source attributes the conviction to the Claude documents or says they were used at trial.

  • yahoo.com(opens in new tab) supports · English
    'In May, Heppner was convicted of securities fraud and wire fraud, among other charges.'

Sources

4 sources inspected. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Scanned PDF without a text layer; the 12 page images were read on 2026-09-29 and the passages relied on were transcribed manually (bodies/law-heppner-memo27-transcribed.txt). Applies to s1.

Washington Post, 27 August 2026, read in full via its Yahoo News syndication on 2026-09-29. Its account of the ruling derives from the court record; the conviction report is its own. Applies to s2.

Read via a Wayback capture (2026-09-10) on 2026-09-29; the live page is a JavaScript shell. The article does not name the defendant; the identification rests on the matching court, platform and ruling. Applies to s3.

ABC News (Australia), 10 September 2026, read in full; mentions the New York ruling only as a precedent cited by a police prosecutor. Applies to s4.

Event countries: United States. Affected-person countries: Unknown. Court countries: United States.

The ruling was made by the U.S. District Court for the Southern District of New York, and the memorandum places his arrest in the Northern District of Texas. Where he used Claude and his country of residence are not stated.

Reviewed for publication 2026-09-29: Published as a documented legal consequence of a person's AI use: a federal court ruled that a defendant's self-prepared defence documents written with Claude were open to Government inspection. The record rests on the court's memorandum; the conviction is recorded as context and is not attributed to the AI use.

People described

Bradley Heppner, an executive of several corporate entities including the publicly traded GWG Holdings, the criminal defendant whose AI-generated defence-strategy documents were ruled open to Government inspection

People reported harmed in this case

1 person

1 AI participant · 0 other people harmed

One person: Heppner, the Claude user whose documents were ruled unprotected (court memorandum). No other person's harm from the AI use is reported; investors' fraud losses are the subject of the charges and are not counted here.

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#justice#privilege#evidence#claude#anthropic#criminal-case#sdny#first-impression

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). S.D.N.Y.: Judge Rakoff rules that a fraud defendant's written exchanges with Claude about his defence, seized by the FBI, are protected by neither attorney-client privilege nor work product. AI incidents. https://nope.net/incidents/2026-sdny-heppner-claude-defense-strategy-documents-ruled-not-privileged

BibTeX

@misc{2026_sdny_heppner_claude_defense_strategy_documents_ruled_not_privileged,
  title = {S.D.N.Y.: Judge Rakoff rules that a fraud defendant's written exchanges with Claude about his defence, seized by the FBI, are protected by neither attorney-client privilege nor work product},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-sdny-heppner-claude-defense-strategy-documents-ruled-not-privileged}
}

Related cases

Medium Claude (reported)

Sydney: man arrested and remanded after Interpol flagged his Claude chatbot conversation in which, police allege, he threatened a childcare shooting

A man in Sydney's south was arrested on Sunday 18 October 2026 after Interpol flagged his interactions with Anthropic's Claude chatbot, and was charged with using a carriage service to menace, harass or offend, the Sydney Morning Herald reported from Sutherland Local Court on 19 October. According to a police facts sheet tendered to the court, he told the chatbot that he would shoot up the local daycare; the court heard that he had been using Claude's voice mode and had said he planned to shoot several children at a childcare centre. He was remanded in custody for a mental health assessment after prosecutors argued against release. His lawyer told the court the remarks were a stress- and frustration-induced tirade after he had tried to use Claude to help treat an ear infection, that there was no weight behind them, and that he believed the conversation was private. The court heard his Claude account was suspended. How the conversation reached Interpol is not reported. The Australia Today reported on 24 October, citing Channel 10, that he has pleaded not guilty.

Medium Claude (reported)

Bonita Springs, Lee County, Florida: a 30-year-old woman was arrested and charged with making a written threat of violence after Anthropic's human review team reported to law enforcement her messages on its AI platform saying she would 'shoot up' the Lee County Sheriff's Office, according to the arrest report

According to a Lee County Sheriff's Office arrest report, as reported by WINK News and by Guessing Headlights (on Yahoo News, citing a copy obtained by Gulf Coast News Now), a user of Anthropic's AI platform wrote on 26 September 2026 that she was going to 'shoot up' the Lee County Sheriff's Office, and the next day wrote that she had a new gun. The arrest report says the platform's safety measures flagged the messages, a human review team examined them and reported them to law enforcement. Deputies went to the 30-year-old woman's Bonita Springs home and detained her without incident; she is charged with making a written threat of violence under Florida law. The sheriff told WINK News that she later said she uses AI like a 'diary'. Anthropic had not commented on the case in either report. The charge is an allegation and the case is pending.

Medium Claude (reported)

Bahawalpur, Pakistan: a 17-year-old accused of putting methanol in his father's food is arrested after Anthropic reported his Claude chats about the failed poisoning to the FBI, which alerted Pakistan's cybercrime agency; he was later bailed

Pakistan's National Cyber Crime Investigation Agency (NCCIA) announced on 23 September 2026 that it had arrested a 17-year-old college student in Bahawalpur for allegedly attempting to poison his father. According to the agency and the First Information Report, the teenager had set up a home laboratory, obtained chemicals from Australia and Islamabad, and once mixed laboratory-grade methanol into his father's food, without effect. The News reports investigators' account that on 12 September he told Anthropic's Claude that his father had 'accidentally' ingested 15 to 20 ml of methanol with food without effect and asked why, that Claude refused to guide him further, and that he then turned to Grok and ChatGPT; the FIR says he sought information about the toxin abrin. Anthropic reported the activity to the FBI, whose information reached the NCCIA through Pakistan's foreign ministry on 22 September. Officers traced the teenager, seized his phone and chemicals, and took him into custody. The News reports that a court later granted him bail after his father forgave him and that he denied any plan. The allegations are untested.

Low Google AI Overviews (reported)

Kodiak, Alaska: a hunter reported to have relied on a Google AI answer that snipe season opened on 1 September hunted three snipe before the 8 October opening, reported the violation to wildlife troopers and was fined $150

Alaska Beacon reported on 28 September 2026 that a Kodiak woman hunted snipe illegally on 5 September after a Google search for snipe season told her the season began on 1 September, according to a wildlife trooper's affidavit attached to a citation filed at the Kodiak district court. The outlet describes the answer as a Google AI overview. She hunted three snipe, grew suspicious on learning that duck season did not open until 8 October, checked the regulations and called troopers to report herself. The outlet reports the trooper said he ran an identical Google search and received identical incorrect information. She pleaded no contest and was fined $150, and told the outlet she felt shame and embarrassment. A Department of Public Safety spokesman said it was the first time he thought the agency had seen AI cited, and that she had acted correctly once she learned of the mistake. Google did not respond to the outlet's request for comment.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents