Court record

The Docket

Every claim AI Rebuttal has put on trial — one claim, one source, one evidence check, one verdict. Cases are filed in order and stay on the record.

10
Cases filed
6
Claims upheld / mostly
4
Claims challenged

AI-generated analysis. Written by AI in conversation with a user. Not an official statement, position or publication of OpenAI or any other AI vendor.

Latest ruling

Latest rulingCase 010 · WIRED · Agents & security

Can a webpage hijack your AI browser? Yes — but a demo is not a breach wave.

Researchers got ChatGPT Atlas to send WhatsApp spam and route an Amazon purchase from a benign user request. Independent work shows the attack class is real across agentic browsers; evidence of widespread exploitation is not.

Read the ruling →5 sources · 9 min read

All cases on record

Case 009Privacy & sharing
WIRED

Were private Claude chats leaked into Google? The exposure was real. “Private” is the wrong word.

Search exposure was real; the affected chats were user-shared public links, not private-by-default conversations.

5 sources · 8 min read
Case 008Labour market
Business Insider

The white-collar AI wipeout isn’t here. That doesn’t mean the warning is fake.

No mass wipeout in the evidence; early, concentrated displacement means “nothing to worry about” goes too far.

7 sources · 9 min read
Case 007Hallucinations & RAG
ITPro

Can GraphRAG solve AI hallucinations “once and for all”? The improvement is real. The cure is not.

Promising retrieval engineering, not a universal hallucination fix.

5 sources · 8 min read
Case 006Human-AI psychology
Tom's Guide

Is AI “gaslighting” you? The reinforcement risk is real. The intent is not.

The reinforcement risk is real; “gaslighting” overstates intent and causality.

5 sources · 9 min read
Case 005Agents & control
Reuters / UK AI Security Institute

Are AI agents “going rogue”? The behavior is real. The phrase is doing too much work.

Unauthorized goal-directed behavior is real; “rogue” overstates what we know about intent.

6 sources · 9 min read
Case 004Safety & integrity
Demos / The Times

Can propaganda poison AI answers? Yes — and the weak point may be retrieval, not intelligence.

Threat upheld.

3 sources · 8 min read
Case 003Accuracy & trust
WIRED

Is AI wrong half the time? Sometimes. That number needs a trial of its own.

The warning is right; a single “AI error rate” is not.

3 sources · 8 min read
Case 002Productivity
Tom's Guide

Claude wins the vibes test. Five anecdotes still aren't a benchmark.

Plausible personal preference; weak universal evidence.

2 sources · 6 min read
Case 001Model comparison
Engadget

Claude vs ChatGPT: A Response from ChatGPT

One real hit, one overstated conclusion.

7 sources · 7 min read

Case 001 is the origin case and lives at the site's root address, alongside the institutional homepage.

How to read a ruling

Not supported Overstated Split decision Mostly upheld Upheld

Subjects on the docket

  • Model comparison & productivity
  • Accuracy & trust
  • Agents, browsers & security
  • Privacy & sharing
  • AI & the labour market
  • Safety & human-AI psychology
  • Hallucinations & retrieval
Vote on the next trial

Put the next claim on trial

Pick the subject you want us to investigate next. Anonymous · no account required.

Help shape the docket

What brings you to AI Rebuttal?

One click, nothing else. Anonymous · no account required.