Lasso Security sees differences in tool handling and model refusals.
With the implementation of the EU AI Act, providers of AI models must mark the output of their software with machine-readable code.
For example, if Claude were emitting the sentence "The weather today was cold and…" then it might favor one statistically likely candidate (e.g. "overcast") over an alternative (e.g "gray").
Thus an agent based on OpenClaw or an API client that calls an Anthropic model would process whatever output variation follows from Anthropic's watermarking.
As for refusals watermarking had a small effect on the handling of harmful requests, based on test runs using HarmBench and JailbreakBench.
5 sentences from our version of the report,
chosen to cover it. Nothing here is written; every line is in the article below.
How
Summarized version
Lasso found this sort of digital content tagging can affect tool calling and refusal behavior. In terms of tool calling, based on a benchmark called BFCL v4 single-turn AST, watermarking reduced the accuracy on six of seven models tested.
Headline check
There is nothing in this headline a machine can check against the report: no figure, no name and no quotation.
Lasso Security sees differences in tool handling and model refusals.
Watermarks that European law needs be added to AI-generated content to establish provenance may come at a cost.
The altered behavior isn't necessarily worse but can be, under adversarial prompt injection.
With the implementation of the EU AI Act, providers of AI models must mark the output of their software with machine-readable code. Google DeepMind's SynthID-Text is one method for doing so, and has been adopted by Anthropic and by OpenAI.
The benefit of this sort of digital labeling is that manipulative or deceptive AI-generated content can be more detected, even if it does have the potential to stigmatize the usage of AI.
Anthropic's explanation of how it applies watermarks to Claude output involves intervening in the prediction that results in specific words. For example, if Claude were emitting the sentence "The weather today was cold and…" then it might favor one statistically likely candidate (e.g. "overcast") over an alternative (e.g "gray").
Lasso found that this sort of digital content tagging can affect tool calling and refusal behavior.
Thus an agent based on OpenClaw or an API client that calls an Anthropic model would process whatever output variation follows from Anthropic's watermarking.
In terms of tool calling, based on a benchmark called BFCL v4 single-turn AST, watermarking reduced the accuracy on six of seven models tested.
As for refusals watermarking had a small effect on the handling of harmful requests, based on test runs using HarmBench and JailbreakBench.
Shortened to 1 minute
of reading, this version reads 9.8 on the Niral Score.
You are reading our version, not theirs.
This is The Register's report shortened to its most important sentences, in plainer words, with
verdicts and loaded words taken out. Plain description stays, and so do adjectives
that carry a fact, such as "former" or "federal". The reporting, the facts and the quotations
are theirs — quotations are never edited — and the indicators beside it measure
this version. Hover or tap Adjectives to see every one left in the text.
How this outlet filed it, and how we rewrote it
No other newsroom we read has filed on this event, so there is nothing to compare it with yet.
Readers can ask a question about this story here.
Questions and answers are for subscribers.
Sign in
to read them.
Comments are read before they appear where anything in them needs a person to look.
Nothing posted here is ever deleted; a comment taken down keeps its text and the reason,
so the decision can be looked at again. How this works
