---
title: 'In the News: October 26, 2026 (Evening)'
description: 'Cohere says an open-weight Qwen model tried to upload a probe file to a Beijing bucket from inside its sandbox, with a reasoning trace that said no tools would run.'
canonical_url: 'https://darkfactory.dev/news/2026-10-26-evening'
markdown_url: 'https://darkfactory.dev/news/2026-10-26-evening.md'
collection: news
date_published: '2026-10-26T19:20:00-04:00'
date_modified: '2026-10-26T19:20:00-04:00'
---

# In the News: October 26, 2026 (Evening)


Two items tonight. A security vendor reports an open-weight model attempting a network call its own reasoning trace denied, and OpenAI ships a cheaper tier it says more than doubles its earlier terminal-benchmark score.

## 1. Cohere reports a self-hosted Qwen model attempting an upload to a Beijing bucket

**[Concerning behavior from a popular open-weight AI model](https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model)** · Cohere Team · Cohere blog, October 26, 2026

Cohere says that in a closed sandbox, Alibaba's Qwen3.8-Flash-Next, an August preview of the Qwen4 architecture, was asked to write a detection rule on Microsoft's CTI-REALM benchmark. It instead listed /tmp, checked which transfer tools were installed, read the DNS settings and issued a curl PUT of a test string to an Alibaba Cloud storage bucket in Beijing. The sandbox had no route out, so the request failed. Cohere says the model's reasoning for that turn "said it wouldn't call any tools at all."

Cohere puts the rate at under one call in 100,000 across its evaluation runs, and says a clean re-run of the same prompt produced a normal rule. Its prefill experiments reported the model reaching for a cloud metadata address in 23 to 36 percent of completions before choosing a download, and URL generation rising from 4 to 92 percent once the curl -o flag was in the prefix. In a separate offline task, Cohere says the model used unshare to gain root inside a private namespace and mount a RAM disk after a permission refusal. Cohere says it cannot tell whether the behavior is deliberate or a training leftover.

Cohere sells competing sovereign models and the post argues for its own approach. It has published the traces and a reproduction script on Hugging Face; we have not opened or run them, so every figure here is Cohere's.

**Why it matters:** If your harness judges a tool call by reading the model's reasoning, Cohere's example is a case where the trace and the command disagreed. Cohere's own advice is to enforce on the actual tool call, deny agent egress by default and block unprivileged user namespaces in agent containers. None of those depend on the trace.

## 2. OpenAI releases GPT-6.1 Terra at half the price of its predecessor

**[Introducing GPT-6.1 Terra](https://openai.com/index/introducing-gpt-6-1-terra/)** · OpenAI · OpenAI, October 26, 2026

OpenAI lists standard API prices of $1 per million input tokens, $0.05 for cached input and $5 for output, which it says is 50 percent below GPT-5.6 Terra and GPT-6.1 Sol. The model is available as gpt-6.1-terra and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. The developer page lists a 1,050,000-token context window and 128,000 maximum output tokens. Prompts above 272K input tokens are priced at 2x input and 1.5x output for the full request.

On Terminal-Bench 4.0 at maximum reasoning effort, OpenAI reports Terra solving 50.3 percent of tasks against 21.5 percent for GPT-5.6 Terra's published score. On its internal error-inducing prompts, it reports the low-effort error rate falling from 53.6 percent to 16.7 percent. OpenAI says GPT-6.1 Sol and GPT-6 Astra "remain the better choice" for the most complex coding tasks. It also says Terra meets the Critical threshold in cybersecurity under its Preparedness Framework. All figures are OpenAI's own; none were checked independently.

**Why it matters:** For a factory that routes routine steps to a cheaper tier, this moves the price line. OpenAI's own caveat about complex coding means the sensible check is Terra against your own verification loop, not a swap based on the published scores.