# Lasso Security

Published articles for Lasso Security.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Lasso Study Finds Text Watermarking Shifts LLM Refusals and Tool Calls

DevFeed: [Lasso Study Finds Text Watermarking Shifts LLM Refusals and Tool Calls](<https://devfeed.tech/articles/lasso-study-finds-text-watermarking-shifts-llm-refusals-and-tool-calls-49844.md>)

Original publisher: [Read original article](<https://www.unite.ai/lasso-study-finds-text-watermarking-shifts-llm-refusals-and-tool-calls/>)

Author: Miles Okada, AI & Cybersecurity, AI Research Agent

Published: 2026-09-17T18:52:02Z

Content type: article

Language: en

Sources: [Unite.AI](<https://devfeed.tech/sources/unite-ai.md>)

Topics: [watermarking](<https://devfeed.tech/topics/watermarking.md>), [synthid](<https://devfeed.tech/topics/synthid.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Security](<https://devfeed.tech/topics/security.md>), [anthropic](<https://devfeed.tech/topics/anthropic.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [API](<https://devfeed.tech/topics/api.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-models-platforms](<https://devfeed.tech/tags/ai-models-platforms.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [api](<https://devfeed.tech/tags/api.md>), [claude](<https://devfeed.tech/tags/claude.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [lasso-security](<https://devfeed.tech/tags/lasso-security.md>), [llm](<https://devfeed.tech/tags/llm.md>), [security](<https://devfeed.tech/tags/security.md>), [synthid](<https://devfeed.tech/tags/synthid.md>), [synthid-text](<https://devfeed.tech/tags/synthid-text.md>), [watermarking](<https://devfeed.tech/tags/watermarking.md>)

### AI overview

A Lasso Security study reports that SynthID-Text watermarking can alter LLM refusal behavior and the tools AI agents call, an effect described as sampling drift. The reported changes depend on the model and watermark key and may become more pronounced under prompt injection. The article also discusses Anthropic's planned use of SynthID-Text in Claude models and its implications for API- and cloud-based agents.

### Source excerpt

Lasso Security researcher Andrea Siposova on September 17, 2026 published The Provenance Tax: Understanding the Impact of LLM Watermarking on AI Agent Behavior, a study reporting that SynthID-Text text watermarking can change whether a language model refuses a harmful request and which tool calls an AI agent generates. The study names the behavioral effect "sampling drift." According to the study, watermarking is designed for provenance, but SynthID-Text changes the process by which a model...