I never claimed the LLMs themselves couldn't "accidentally" leak the information through. Obviously, the LLMs can do anything they want.
The issue here is whether the harness is doing it. Claude Code is doing it. Codex is not.
The problem here literally is the non-deterministic model cooking up commands with the user email address in the header, and the harness not blocking that (and the implemented fix is to deterministically block that), not the harness deterministically hardcoding the email in the web search tool. No technical report I’ve seen has claimed the latter, although this particular linked report which is low quality and vague can be misinterpreted as that.
> although this particular linked report which is low quality and vague can be misinterpreted as that
Seems that way... Here's how I interpreted it:
> claude code divulged my email to arbitrary 3rd parties in the User-Agent request header while conducting web searches without my permission
"Claude Code" refers to the harness, not the model. Web search is also one of the tools provided to the model by the harness.
https://code.claude.com/docs/en/tools-reference#websearch-to...
https://platform.claude.com/docs/en/agents-and-tools/tool-us...
loading story #50386633