Safeguard
AI Security•

Using a personal AI chat account for legal matters: the risk

Anthropic's help center cites a 2026 ruling where a defendant's consumer AI use was not protected work product. What it means for personal chats.

Safeguard Security Team
3 min read

If you are in a legal dispute, think twice before working through it in a personal AI chat. Anthropic's help center cites a February 2026 federal ruling where one defendant's consumer AI use was not protected work product. It is one non-appellate decision, so talk to a lawyer about your own case.

What the help center says

Anthropic published an article on using Claude for legal work, covering privilege, confidentiality and configuration. It describes a case, United States v. Heppner (S.D.N.Y., February 2026), in which a criminal defendant's own use of a consumer AI account, without his lawyers' involvement, was not protected work product.

The article says the court's reasoning relied on the consumer privacy terms and the lack of attorney direction. It also notes this is not an appellate decision, and courts elsewhere may differ. It also says that keeping a lawyer directly involved strengthens privilege and work product protection.

This is not legal advice, and neither is this post.

Why the account type matters

The same article draws a line between product types. It says Anthropic does not train on your content by default for commercial offerings: Claude for Work, the Claude Platform and the API. It does not describe training terms for consumer accounts, and says to check those separately.

Retention controls differ too. Administrators on work plans can set organization retention, and API customers get standard 30 day retention or zero data retention on eligible endpoints. Those were not described as controls for individual consumers.

Practical habits for individuals

  1. Do not paste a letter from a lawyer, a court filing or a settlement offer into a consumer chat.
  2. If you need help understanding a legal document, ask your lawyer, or ask general questions without personal details.
  3. Never describe an incident you are involved in, in detail, to a chat tool. The text may be retained, may be reviewed if flagged, and can be requested in some proceedings.
  4. If a lawyer asks you to use an AI tool, get the instructions in writing and use the account they tell you to use.
  5. Delete is not a plan. Deleting a chat from your history does not undo having written it.

For small businesses

If your team uses AI for contracts, HR matters or disputes, choose a business plan with admin controls and write a short rule about what staff may paste in. Our post on what to keep out of AI chats gives a starting list.

Quick check

Does one court decision settle this?

No. Anthropic's article says it is not an appellate decision and other courts may differ.

Is a work account automatically privileged?

No. The article says involving a lawyer directly strengthens protection. The account type alone does not decide it.

Sources

Last reviewed: 19 October 2026.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.