ChatGPT draws the Gruffalo in under a minute, and its illustrator, Axel Scheffler, is weighing a lawsuit. If you make images for clients, another artist’s character is not yours because a prompt produced it. Also: DataDome’s agent traffic report, Meta’s new crawling terms and the brakes Brad Smith wants.
The Gruffalo in under a minute
Illustration generated with AI on our own machine.
What happened. Axel Scheffler, the illustrator of the Gruffalo, is considering joining a legal fight against major tech companies, heise reports. Asked for an image of the Gruffalo, ChatGPT produces results in under a minute. Penguin Random House is the first book publisher in Europe to take legal action against OpenAI over alleged copyright infringement. The Carlsen publishing house, author Marc-Uwe Kling and illustrator Astrid Henn are also suing OpenAI.
What we think. We follow these cases because they decide who answers for a generated image; you, if you sell illustrations, do not deliver a character you did not create. One allegation is that ChatGPT offers print-ready layouts with a fake ISBN and a publisher’s logo. OpenAI’s answer is that AI does not store data in the conventional sense.
The good bit. A major publisher has taken it to court, so there will be a legal answer rather than more opinions.
What you can do. If you use AI for client images, add a line about third-party characters to the brief, and check what the EU AI Act guide asks you to document.
25 billion agent requests
Illustration generated with AI on our own machine.
What happened. AI agent traffic across DataDome’s network grew 41.5% quarter on quarter to nearly 25 billion requests in Q3 2026, Help Net Security reports. Meta’s Muse generated 29.1 million requests between 22 September and 3 October. In a sample, 8.9% of requests reached sensitive endpoints, and forms took 81% of that traffic.
What we think. We watch the forms before the login page, because that is where the agents go; you, if you run a contact form, check how many submissions come from bots. Forms received about eight times as many requests as login pages. In a sample of Muse traffic, 99.3% of requests came from cloud and CDN infrastructure.
The good bit. A report broken down by agent and by page type tells you where to put the filter, without guessing.
What you can do. Add a per-IP submission limit to your form this week, and if you want to filter on your own machine, see the local AI developer setup.
Meta deletes the word train
Illustration generated with AI on our own machine.
What happened. Meta revised its commercial terms after wording about website crawling raised concerns about training its AI models on business websites, Digital Trends reports. The new version removes the word “train” and limits collected material to Meta’s advertising and business products. An Axios analysis of 102 major U.S. publishers found that 42% block AI training crawlers.
What we think. We read which verbs are left: “research” and “improve” are still there; you, if you run a business website, decide in your robots.txt rather than in their terms. Removing a word does not remove what was already crawled. Meta still uses public Facebook and Instagram posts for training in some regions.
The good bit. Meta confirms site owners can keep setting their crawling preferences through robots.txt.
What you can do. Open your robots.txt and add rules for Meta’s crawlers; if you handle customer data, also read why GDPR and local deployment fit together.
A washing machine has more controls
Illustration generated with AI on our own machine.
What happened. Microsoft President Brad Smith warned about the lack of safety controls in advanced AI systems in a BBC interview broadcast on 12 October, Business Insider reports. Smith proposes emergency brakes to stop AI agents that act unexpectedly. He compared AI’s safety controls with those of a washing machine or a milk carton.
What we think. We put a person between the agent and any action that cannot be undone; you, if you are trying agents, start with read-only tasks. When the president of a company that sells agents asks for brakes, that says enough about the brakes.
The good bit. The EU AI Act has been in force since 1 August 2024, and the Spanish Government approved a draft Organic Law on AI governance in May 2026.
What you can do. List the actions your agent can take and mark the ones that need your approval; the EU AI Act guide tells you what to record.
In one line
If ChatGPT draws the Gruffalo in a minute, Scheffler’s question is not technical: it is who pays when the drawing is sold.
Work with us
If something in this column belongs in your business, on models that run on your own hardware rather than someone else’s cloud, tell us what you need or see how we work in consulting.