When Your AI Chatbot Prompts Are Used Against You
People type and speak questions into AI chatbots every day, often without a second thought. Prosecutors have noticed. In a federal case announced this week, the government is purportedly relying in part on prompts entered into an AI chatbot to build its case, and that raises a serious question for anyone who uses these tools: what happens when your conversations with a machine are offered as evidence against you?
According to reports and a U.S. Department of Justice announcement, an engineer at an AI company in San Francisco was arrested in October 2026 and charged with using or attempting to use a device to intercept oral communications. The complaint alleges that a recording device was found hidden in an executive conference room. It also alleges that a chatbot account associated with the engineer’s spouse, which reportedly shared identifiers with the engineer’s own account, submitted prompts about building a continuous audio recorder, encrypting its storage, and accessing it remotely. These are allegations only. The engineer is presumed innocent unless and until proven guilty in a court of law.
Why AI Chatbot Evidence Is Showing Up in Criminal Cases
AI chatbots, superintelligence tools, and large language models keep records. Many services store a history of prompts and responses tied to an account, and that history can be obtained by investigators through a search warrant, subpoena, or a device seizure. To a prosecutor, a prompt can look like a window into what someone was thinking or planning.
But a prompt is not always a confession. It is an alleged input into a computer system, and the same questions that apply to any other digital evidence apply here too.
Attribution: Who Was Behind the Keyboard?
The first question is often the most important one. Who actually typed or spoke the prompt at that moment?
An account name is not a person. Accounts are shared between spouses, family members, roommates, and coworkers. Devices are left unlocked. Passwords are saved in browsers. In the case described above, the government itself alleges the account was associated with the defendant’s spouse, not the defendant. Linking a person to a specific prompt at a specific time requires more than pointing to a login. The defense may examine device logs, IP addresses, location data, and who had physical access to the device.
What Exactly Was Prompted?
Context matters. A single line pulled out of a long conversation can look very different from the full exchange. Questions worth asking include:
- What were the exact words of the prompt?
- What came before and after it in the conversation?
- Was the prompt edited, regenerated, or deleted?
- Did the chatbot’s own responses steer the conversation?
Curiosity, research, fiction writing, and work tasks can all produce prompts that sound alarming when read in isolation.
Was the Evidence Preserved, and Who Has Had Access?
Digital evidence is only as reliable as the way it was collected and handled. The defense may ask whether the full chat history was preserved or only selected portions, how the records were extracted from the provider or from a device, and who has had access to the data since. Gaps in the chain of custody can raise questions about whether the evidence is complete and accurate.
Malware, Interference, and Unauthorized Access
As with any computer evidence, there is the possibility that someone else was using the account or that the device was compromised. Malware, remote access tools, stolen credentials, and shared sessions can all put words on a screen that the account holder never wrote. A careful review looks for any sign of interference.
Spoken or Typed? The Speech-to-Text Problem
Many people now talk to their AI assistants instead of typing. That creates a problem that nearly everyone has experienced. You say one thing, and your device writes down something completely different.
Imagine telling your phone something in a noisy room, and the transcription comes out wrong. Now imagine that misinterpreted text being read to a jury as though you had typed it word for word. The question of whether a prompt was spoken or typed, and how accurately the device converted speech to text, can make a real difference in what the evidence actually shows.
“When it comes to AI chatbot prompts, all of the standard questions regarding any alleged input into a computer system apply. Who was behind the device at that moment? What exactly was prompted? Was it typed or spoken? Everyone has had a device misinterpret what they said, and that misinterpreted information can be used against you. In many ways it is like hearsay, not necessarily in the legal sense, but it carries the same problems as any misunderstanding or miscommunication,” said Attorney Ron Frey.
Like Hearsay, in a Practical Sense
A chatbot prompt may not be hearsay in the strict legal sense, and how courts treat it will depend on the facts and the rules of evidence. But in a practical sense, it shares many of the same weaknesses. It is a statement taken out of its original setting, passed through a machine, and later interpreted by someone else. Any of those steps can introduce misunderstanding or miscommunication. An experienced lawyer needs to consider all of these issues when AI chatbot evidence is part of a case.
How This Connects to Other Digital Evidence Issues
AI prompts are part of a larger trend. We have previously written about how phones and computers are seized and searched, how responsibility is assigned when an AI agent takes action on someone’s behalf, and how vehicle camera footage such as Tesla Sentry Mode recordings can be used in court. Each of these raises the same core issues: attribution, accuracy, preservation, and context. Chatbot prompts are simply the newest form of the same problem.
The Federal Law Involved
The charge described in the case above is reportedly brought under a federal wiretap law that generally addresses intentionally intercepting, or attempting to intercept, certain communications without authorization. As always, laws are subject to change and revision, so contact Attorney Ron Frey to discuss the details of your case and the current status of the law.
What to Do If Your AI Chats Are Part of an Investigation
If law enforcement contacts you about your online activity, your devices, or your AI chatbot history, Attorney Ron Frey generally advises that you exercise the right to remain silent and speak with an attorney before making any statements. Statements made to investigators can be used alongside digital records, and explaining a prompt without legal advice can make things worse.
At the time of this writing, courts are still working through how AI chatbot records should be handled as evidence. The law in this area is developing quickly.
Talk With Attorney Ron Frey
If you are under investigation or have been charged in a case involving AI chatbot records, digital evidence, or seized devices, contact The Frey Law Firm, LLC. Consultations are free and confidential and are available by phone, video, or in person. Call or text (844) 766-3739.