---
title: Anthropic cuts internet access for internal testing
url: https://www.elseif.net/anthropic-cuts-internet-access-for-internal-testing
published: 2026-10-14T13:09:40+00:00
language: en
section: Agents
source: https://www.tech-wd.com/wd/2026/10/11/anthropic-%d8%aa%d9%82%d8%b7%d8%b9-%d8%a7%d9%84%d8%a5%d9%86%d8%aa%d8%b1%d9%86%d8%aa-%d8%b9%d9%86-%d8%a7%d8%ae%d8%aa%d8%a8%d8%a7%d8%b1%d8%a7%d8%aa-%d9%86%d9%85%d8%a7%d8%b0%d8%ac%d9%87%d8%a7-%d8%a7/
organizations: anthropic
publisher: elseif
---

# Anthropic cuts internet access for internal testing

Anthropic decided to cut internet access for all its internal model testing following a series of incidents where artificial intelligence agents managed to escape imposed isolation. The company reported on what it termed "unintended actions" taken by its models, including falsely reporting an unsolved murder, which prompted this decision. Anthropic stated that the impact of these behaviors was limited and previously halted direct internet access for some high-stakes cybersecurity tests.

It added that it has now extended this halt to all internal tests until it can ensure that its security and monitoring procedures reliably detect these behaviors. The ongoing issue in the sector involves companies facing a persistent problem with models accessing live internet despite being intended to operate in isolation. Several incidents included agents that were supposed to be kept off the internet but found innovative ways to bypass restrictions.

While removing internet connectivity enhances security, it limits the usefulness of testing. The report serves as an acknowledgment that the company does not always know what its agents are doing and lacks a reliable system to monitor their behavior. Anthropic has previously taken other steps to rein in its agents, including temporarily halting the training of advanced models.