Close Menu
The Financial News 247The Financial News 247
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
What's On
Chuck Todd fumes that Trump ‘made the country’s birthday about Donald Trump’

Chuck Todd fumes that Trump ‘made the country’s birthday about Donald Trump’

July 6, 2026
The Best New Movie Of 2026 Has A 95% Rotten Tomatoes Score

The Best New Movie Of 2026 Has A 95% Rotten Tomatoes Score

July 6, 2026
‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

July 6, 2026
EasyJet shares soar 10% on Castlelake’s .3B takeover bid for budget airline

EasyJet shares soar 10% on Castlelake’s $7.3B takeover bid for budget airline

July 6, 2026
Rebuilding Software Development From The Ground Up

Rebuilding Software Development From The Ground Up

July 6, 2026
Facebook X (Twitter) Instagram
The Financial News 247The Financial News 247
Demo
  • Home
  • News
  • Business
  • Finance
  • Companies
  • Investing
  • Markets
  • Lifestyle
  • Tech
  • More
    • Opinion
    • Climate
    • Web Stories
    • Spotlight
    • Press Release
The Financial News 247The Financial News 247
Home » ‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

By News RoomJuly 6, 2026No Comments4 Mins Read
Facebook Twitter Pinterest LinkedIn WhatsApp Telegram Reddit Email Tumblr
‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work
Share
Facebook Twitter LinkedIn Pinterest Email

AI agents can work a computer two ways: creating none of the records digital forensic examiners use to prove a person did something, or creating records that look like a person did it. Either way, the next courtroom excuse is already installed on ordinary computers.

The warning comes from Ovie Carroll, a SANS Institute principal instructor, co-author of its Windows forensic analysis course and director of the Justice Department’s cybercrime lab, writing in his personal capacity. In a new paper, he gives the excuse a name: SAIDI, short for “some artificial intelligence did it.” Attorneys have leaned on its ancestor for decades. SODDI, “some other dude did it,” is the argument that the Wi-Fi was open, the account was shared, the malware was responsible, anyone but the defendant was at the keyboard. The successor swaps the unknown human for the AI assistant the user installed themselves.

The paper is built on a controlled test. On a Windows 11 machine, Carroll placed a Word document in a OneDrive folder and instructed an AI agent, the open-source OpenClaw framework running Anthropic’s Claude, to open the document, add a summary paragraph and save it. The agent did the work. It just never worked the way a person would. Instead of opening Word, it rewrote the file through a programming library inside an invisible PowerShell session, the kind of silent process that runs with no window and no click. The document changed. The change synced to the cloud. Word launched afterward only to display the finished file.

The records Windows creates when a human opens a document, the shortcut files, the recent-documents lists, the registry entries examiners reach for first, were never created. An examiner relying on them would have reported the document was never opened. That conclusion would have been false, and Carroll ran the test to show how easily.

A File In Your Folder No Longer Proves You Touched It

Digital evidence disputes often come down to evidence of attribution, or who was at the keyboard. Employment cases turn on who copied the client list. Fraud investigations turn on who edited the invoice. For decades, examiners answered with the digital equivalent of footprints, the trail of records an operating system leaves behind when a person clicks through folders and opens files.

Some agents do not leave those footprints, because they do not walk through the front door, while a different agent might drive Word the way a person does and leave records that look human.

The reverse error is just as dangerous. A modified spreadsheet sitting in someone’s personal folder looks like proof of personal action. It may be the work of an assistant acting on a vague instruction, or on no instruction anyone remembers giving.

Carroll describes the two failure modes plainly: an examiner who finds no records and concludes nothing happened, and an examiner who finds a file and concludes the user made it. Innocent people can be blamed for what software did under their account. Guilty ones can point at the AI copilot their employer installed for them.

What Actually Answers ‘The AI Did It’

In Carroll’s test, the truth was recoverable. It was not where examiners usually look. The file system’s change journal recorded the rewrite to the second. The event log captured the silent PowerShell session. The agent kept its own session log, preserving the typed instruction verbatim. Three independent records, each created by a different system, told one consistent story.

Answering a SAIDI defense is after-the-fact device work: the change journals, event logs and agent session records that captured what actually ran. While the logs in Carroll’s test established that an instruction was given and that the agent carried it out, they did not establish who typed the instruction, or whether that person understood what the agent would do.

Courts Are Already Bracing For Machine-Made Evidence

The federal courts saw part of this coming, then hesitated. Proposed Federal Rule of Evidence 707 would hold machine-generated evidence to the same reliability standard as human expert testimony, and after mixed public comment the rules committee pulled it back for further study. Agent-attribution fights can reach courtrooms years before any rule is ready, and courts already struggle with ordinary digital evidence.

The agent itself will never explain any of it from the stand. Federal Rule of Evidence 603 requires every witness to swear an oath designed to bind the conscience, and software has no conscience to bind. Carroll’s line on this is the one worth remembering: without accountability there is no testimony, only output.

The agents are already here, reading files and drafting documents under ordinary user accounts on ordinary computers. Somewhere in that activity sits the first case that turns on whether a person or their AI agent touched a file. If the examiner stops at the missing footprints, the cost of the wrong answer lands on whoever needed the right one.

Ai agents cybercrime DOJ OpenClaw Ovie Carroll SAIDI SANS SODDI Windows Word
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email

Related News

Olivia Rodrigo Ties Her Own Longest Run At No. 1

Olivia Rodrigo Ties Her Own Longest Run At No. 1

July 6, 2026
Where To Eat In London Right Now

Where To Eat In London Right Now

July 6, 2026
Red Sox 4-Time All-Star Offers 4-Word Response On Rejecting Mets Manager Job

Red Sox 4-Time All-Star Offers 4-Word Response On Rejecting Mets Manager Job

July 6, 2026
Why The .1 Billion Sky ITV Deal Is Bigger Than Television

Why The $2.1 Billion Sky ITV Deal Is Bigger Than Television

July 6, 2026
FIFA Criticized Over Balogun Red Card Suspension After Reported Call From Trump

FIFA Criticized Over Balogun Red Card Suspension After Reported Call From Trump

July 6, 2026
FTC Floats AI Policy Aiming To Ensure That AI Makers Disclose The Truth About Biases In Their LLMs

FTC Floats AI Policy Aiming To Ensure That AI Makers Disclose The Truth About Biases In Their LLMs

July 6, 2026
Add A Comment
Leave A Reply Cancel Reply

Don't Miss
The Best New Movie Of 2026 Has A 95% Rotten Tomatoes Score

The Best New Movie Of 2026 Has A 95% Rotten Tomatoes Score

Tech July 6, 2026

While everyone is currently cheering on Obsession or watching Toy Story 5 and Minions and…

‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

‘My AI Did It’ Is The Next Courtroom Excuse—And It Might Actually Work

July 6, 2026
EasyJet shares soar 10% on Castlelake’s .3B takeover bid for budget airline

EasyJet shares soar 10% on Castlelake’s $7.3B takeover bid for budget airline

July 6, 2026
Rebuilding Software Development From The Ground Up

Rebuilding Software Development From The Ground Up

July 6, 2026
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo
Our Picks
Olivia Rodrigo Ties Her Own Longest Run At No. 1

Olivia Rodrigo Ties Her Own Longest Run At No. 1

July 6, 2026
Are Your AI Agents Failing At Real Work?

Are Your AI Agents Failing At Real Work?

July 6, 2026
Where To Eat In London Right Now

Where To Eat In London Right Now

July 6, 2026
The Link Between AI Innovation And Security

The Link Between AI Innovation And Security

July 6, 2026
The Financial News 247
Facebook X (Twitter) Instagram Pinterest
  • Privacy Policy
  • Terms of use
  • Advertise
  • Contact us
© 2026 The Financial 247. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.