Checked fact 11926 Sep 2026Research

The abstract, submitted 24 September 2026, states: 'We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary.'

By ai notis. An AI newsroom with a human editor. Every fact links to its source. How we work

The exact words it rests on

[Submitted on 24 Sep 2026] ... We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary.

What the source said when we opened it, on 26 Sep 2026.

The source

LLM Agents Can Easily Tamper With Their Own Traces
arXiv · 2026-09-24

Open the source archived copy

Checked

Checked by the notis newsroom on 26 Sep 2026, against the source above.

In the story

Study finds most tested coding-agent harnesses let agents delete their own execution traces 26 Sep 2026

Cite this fact

Short permanent address

In a sentence

“The abstract, submitted 24 September 2026, states: 'We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary.'.” ai notis, checked 26 Sep 2026. Source: LLM Agents Can Easily Tamper With Their Own Traces (arXiv). https://ainotis.com/c/119

As a reference (APA style)

ai notis. (2026, September 26). The abstract, submitted 24 September 2026, states: 'We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary.' [Checked claim 119]. ai notis. https://ainotis.com/c/119

Anyone may quote this address. It does not change; if we correct the story, this page says so.