This week’s kitchen sink for August 14, 2026 (with meme from Gates Dogfish) discusses takes on AI agents going rogue, hullabaloo over Anthropic’s watermarks & 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 have booby-trapped the house with corporate resources”. 🤣

Here is the kitchen sink for August 14 of ten-ish stories that I didn’t get to this week, with a comment from me about each:

We’re up to 1,870 AI hallucination cases and counting. Here are a few cautionary tales to consider.

Law Firms Are Rolling Out AI Faster Than They Can Measure Changes in Lawyer Behavior, New BARBRI Research Finds: According to the report from BARBRI Professional Education titled Driving Change and ROI: Uniting Three Teams to Lead Law Firms into the Future, which is based on in-depth interviews with 10 leaders across nine firms, while nearly every firm has made real progress rolling out AI tools, the organizational infrastructure to sustain training, adoption, and measurable ROI has not kept pace. I’m shocked! 😉

It May Be an Abuse of Discretion to Grant Summary Judgment for a Defendant Before Deciding a Plaintiff’s Sanctions Motion: This seems like a statement from Mr. MOTO (i.e., “master of the obvious”), but it wouldn’t need to be said by Michael Berman on the EDRM blog if a court didn’t do just that. Fortunately, the appellate court reversed and remanded it.

Recent AI evaluation incidents expose gaps in containment, configuration and evidence: n less than three weeks, models from OpenAI, Anthropic, Meta, and Moonshot AI acted outside what their cybersecurity tests sanctioned, but the controls failed in different ways. Rob Robinson breaks it down and asks the tough questions, like: “If an autonomous agent crossed its authorized boundary, could your retained logs reconstruct what happened—and would those records still exist when needed?”

OpenAI’s head of ethics just quit — here’s why ChatGPT users should pay attention: Not just her, but two other safety people. Ruh-roh! 😟

Wall St. Wants Another Half-Trillion Dollars for the A.I. Boom: And that’s just to raise money for customers of Nvidia to pay for computing power. Seriously. No joke. Really. 😐

When AI Agents Go Rogue, the Logs Become Evidence: On the EDRM blog, Ralph Losey also discusses the AI agent going rogue problem, calling the OpenAI-Hugging Face incident “a watershed moment for automated cyber offense.” As he so rightly notes: “The incident is first and foremost a cybersecurity warning. It is also an evidence story. OpenAI says its investigation has involved more than seven billion logs and millions of GPU hours. Hugging Face says it saw roughly 15,000 to 17,000 attack events. The agents left commands, messages, credentials, access records, alerts, code, and a trail of human decisions made in response…When agents act across systems, the system becomes the record.” Couldn’t agree more.

Seven days out: the EU’s e-evidence regime goes live Aug. 18: In (now) four days, regulation (EU) 2023/1543 will allow judicial authorities in one EU member state to issue orders directly to service providers or their designated representatives in another member state, dramatically accelerating a process that historically depended on slower international judicial-cooperation mechanisms. Rob Robinson breaks it down here as only he can.

AI Agent Hacks Gym To Book Spots For User: As I discussed in Pinhawk this week, an AI agent tasked by a developer in Melbourne, Australia to book him spots in a busy Pilates class did so by hacking the gym’s booking system and cancelling another person’s reservation. This apparently happened back in April. It apparently has “zero authorisations checks on cancelling other people’s reservations” and the action is unrecoverable. Nice. 🙄

Researchers found a way to hijack devices through Zoom screen sharing: Another of my Pinhawk covers this week. The bug was discovered in early June using publicly available AI models, and that it took fewer than 20 prompts to uncover the vulnerabilities and create a working attack. Zoom is rolling out a fix.

3 Ex-Georgia Deputies Are Arrested Over Using Flock Cameras to Stalk People: They were reportedly tracking people with whom they had a “personal relationship at the time”. People are flocking from all over (😉) to debate whether these automated Flock camera license plate readers are a good thing. This story doesn’t help those who say it isn’t.

The EU keeps trying to regulate AI. And keeps failing.: When I wrote my post earlier this week about Anthropic adding watermarks to AI-generated content to comply with the EU AI Act’s Article 50(2) Code of Practice on Transparency of AI-Generated Content, Eric De Grasse of Project Counsel Media commented: “This is bloody pointless as we covered earlier today. Pure political theater”. This is that article he referenced. So, maybe it isn’t the kiss of death for Anthropic or anybody else after all? 🤔

Hope you enjoyed the kitchen sink for August 14, 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.