Your AI Ghostwriter Just Lost Its Anonymity

Something changed this week that touches almost everyone on this list, whether or not you have thought about it yet.

Anthropic, the company behind Claude, the model that runs AskDimitri, announced that Claude models launched on or after August 2, 2026 embed an invisible, machine readable watermark in the text they generate. Files such as images carry a related but separate signed authenticity record. The company says the mark does not change the meaning or quality of the writing, travels with the text when it is copied and pasted, and can survive light editing.

August 2, 2026

The trigger is European regulation, specifically Article 50 of the EU AI Act, which requires generative AI providers to mark synthetic content so it can be identified. Anthropic was among about 190 organisations, including Microsoft, Google, and Meta, that signed the EU's voluntary Code of Practice on Transparency of AI-generated Content, which providers can rely on to show they comply with this rule. Because the marking happens at the model level rather than by region, it applies to everyone using a supported Claude model, everywhere, regardless of where you live.

Now, before this turns into a story that is too neat, let me tell you exactly what it does and does not do.

This is a real step toward transparency, not a solved problem. Treat it as raising the cost of undisclosed AI ghostwriting, not as making it impossible.

Here is why this matters specifically for research integrity, and it connects directly to something I wrote about a few weeks ago on the rise in retractions. Undisclosed AI generated text in manuscripts is an emerging category that journals are actively watching, alongside paper mills and image manipulation. Most major journals, including those bound by ICMJE guidance, already require authors to disclose any generative AI use in preparing a manuscript. A watermark does not replace that requirement. It simply means the gap between what was disclosed and what was actually used is now somewhat more detectable than it was a month ago, at least for supported models and lightly edited text.

The honest reading is not that AI ghostwriting is now impossible to hide. It is that the entire direction of travel, across regulation, journal policy, and now model level infrastructure, points toward disclosure becoming the default expectation rather than an afterthought. Getting ahead of that shift is a better position than being caught behind it.

AskDimitri runs on Claude, which means every conversation already sits on the transparent, disclosed side of this shift. It was built to be the tool you cite, not the one you hide, and that has not changed. If you want help thinking through how to disclose AI assistance properly in your next submission, AskDimitri can walk through the specific language most journals expect.

Get your disclosure language right →

The tools are not going away, and neither is the pressure to use them. The only real choice left is whether you are the researcher who used AI and said so, or the one hoping nobody checks.

Editor's note: in October 2026 the facts in this letter were checked against the sources listed below, and the text was corrected where the letter as sent differed from them.