Watermarking a model's output made it worse at calling tools and weaker under prompt injection
Watermarking is treated as a free addition that only affects provenance. This is the first published claim that it changes what an agent does.
Company
Lasso Security is an organisation working in artificial intelligence. This page collects every Aivio News story about Lasso Security, together with its models and recorded benchmark results.
Watermarking a model's output made it worse at calling tools and weaker under prompt injection
Watermarking is treated as a free addition that only affects provenance. This is the first published claim that it changes what an agent does.
Privacy preferences
We use necessary cookies to run this site. With your consent we also use advertising cookies. You can change your choice at any time. Cookie Policy · Privacy Policy