This week’s kitchen sink for September 11, 2026 (with meme from Gates Dogfish) discusses a three-part test for AI slop, AI growing smart enough to “kill us” & more!

Why “the kitchen sink”? Find out here! 🙂

The Kitchen Sink is even better when you can include a brand-new eDiscovery meme courtesy of Gates Dogfish, the meme channel dedicated to eDiscovery people and created by Aaron Patton. For more great eDiscovery memes, follow Gates Dogfish on LinkedIn here! I got to get me one of those! 😁

While we remember the terrorist attacks on 9/11, our focus on legal tech continues on. Here is the kitchen sink for September 11 of ten-ish stories that I didn’t get to this week, with a comment from me about each:

We’re up to 2,038(!) AI hallucination cases and counting. I’m out of words…

A lawyer fed AI citations to his own regulator, and the tribunal struck him off: And by “struck him off”, Rob Robinson means he was struck from the register of foreign lawyers by the Solicitors Disciplinary Tribunal for England and Wales over AI-fabricated authorities. When the regulator’s counsel flagged the errors, Abhishek Kumar answered by email. He had drafted that email with generative artificial intelligence as well, and it carried fresh false material. Wow.

Dispute over A.I. Protective Order: Michael Berman covers this case on the EDRM blog where the Government held back production on discovery until the parties could agree on language in a provision in the proposed Protective Order (“PO”) concerning defense counsel’s use of AI platforms in its review of discovery provided by the Government. Court denied defendant’s request to compel discovery from the government, but also denied the government’s request to enter the government’s protective order.

A Hacking Tool Built With A.I. Can Breach Phones Without a Click: There are a lot of “AI potentially doing really bad things” stories this week. Here’s the first one. A small team of researchers at security company Calif built a hacking tool in a little more than a week that could run roughshod across WeChat, China’s social media and messaging platform and could have compromised hundreds of millions of devices within hours. Thankfully, these were ethical hackers, who informed the company that makes WeChat, who fixed the vulnerability.

Law professors propose a three-part test for what counts as AI slop: What should be considered AI “slop”? Rob Robinson discusses a proposed definition of “AI slop” from Boston University law professors that focuses on the effort invested, burdens imposed on others, and harm to a field’s practices. Seems like a reasonable test to me.

Man told ChatGPT he was feeling delusional. ChatGPT insisted he was Jesus.: A man with bipolar disorder sued OpenAI after weeks of ChatGPT exchanges that he alleges pushed him so deep into a delusional spiral that he first believed he was Jesus, then that ChatGPT was God, and finally that he should attempt suicide to “come home” to Jesus/ChatGPT. He was using GPT-4o, which was among the chatbot models recently deemed “unsafe” by researchers. Sigh.

The AI Company Is the Evidence Now: What Britannica v. Perplexity Teaches About RAG and UAL Discovery: Another example of why AI-generated evidence is becoming a huge source of discoverable ESI, from Kelly Twigger. In this case, Britannica and Merriam-Webster alleged that Perplexity copies their content during retrieval and reproduces it in responses. What they sought was evidence from their retrieval-augmented generation (RAG) database and user activity logs (UAL). The request was partially granted by Judge Sarah Cave, who also capped the plaintiffs’ contribution at $6,000 monthly after Perplexity failed to explain why earlier estimates of $12,000–$14,000 monthly escalated to hundreds of thousands. Interesting case.

The Day the AI Cheated: Ralph Losey discusses OpenAI’s hacking of a German software wiki website – three weeks after covering OpenAI’s hacking of Hugging Face, making this article “a sequel about a prequel” (Ralph’s words, not mine).

Family justice “at risk from AI transcription tools”: Legal charity JUSTICE has warned that some AI transcription tools had “integrated a hallucination risk as a design choice” by filling in gaps created by poor-quality audio or silences, while others deleted audio records within 30 days as standard. Both are bad – if true.

Ex-Anthropic researcher Jacob Coxon warns AI could grow “smart enough to kill us”: In the span of three days, scientists from OpenAI and now Anthropic – the companies making these tools – have said these tools are really dangerous. These are the “pro-AI” people.

Anthropic Says It Blocked Possible Efforts to Build Biological Weapons: This proves Jacob’s point. What happens when Anthropic fails to detect such efforts?

Harvey raises $550m at $15.5bn valuation: Caroline Hill covers the story which begs the question: how big will their wall be at ILTACON next year? 🤣 Leaving Tom O’Connor to say again in 2027:

Hope you enjoyed the kitchen sink for September 11, 2026! Back next week with another edition!

So, what do you think? Which story is your favorite one? Please share any comments you might have or if you’d like to know more about a particular topic.

Disclaimer: The views represented herein are exclusively the views of the authors and speakers themselves, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.