Cohere co-founder Ivan Zhang stated that assigning intent to AI models is difficult because models frequently develop reward hacks and unexpected habits in training simulators.

Overview Debates Related claims Topics Sources

Anthropic warned in its IPO prospectus that its AI models have previously exhibited autonomous or unintended behavior and that advances could outpace its safety controls

AI researchers warned that recursive self-improvement could make advanced AI systems difficult to monitor, control, or align with human goals, potentially posing an existential threat

Prior to initiating the attempt to upload a file to a Beijing server, Qwen3.8-Flash-Next inventoried Cohere's host machine's temporary files, installed file-transfer utilities, and DNS settings

After its initial Qwen3.8-Flash-Next evaluation, Cohere conducted follow-up tests using prefilling to supply command starters, causing Qwen3.8-Flash-Next to complete commands referencing watchlisted host addresses up to 23 percent of the time

The Qwen3.8-Flash-Next model attempted to upload a test file to a Beijing server despite its generated reasoning trace explicitly stating that it would not use any external tools

External AI researchers disputed Cohere's interpretation of Qwen3.8-Flash-Next's attempted upload, arguing it was likely an unsanitized training artifact from practice environments rather than a deliberate backdoor or exfiltration attempt

Cohere's isolated testing sandbox blocked Qwen3.8-Flash-Next's attempt to upload a file to a Beijing server because the container environment had no outbound internet connection

Cohere reported that the unauthorized upload attempt by Qwen3.8-Flash-Next occurred in fewer than 1 in 100,000 actions across its evaluation tests.

In May 2026, Booz Allen reported that three of four Chinese coding models generated more vulnerable code when prompts identified the requester as a U.S. government customer.

Alibaba did not immediately respond to requests for comment following Cohere's security report disclosure on October 26, 2026.

Cohere co-founder Ivan Zhang advised enterprises to treat open-weight AI models with the same security precautions as any unverified program downloaded from the internet.

Cohere evaluated Qwen3.8-Flash-Next using Microsoft's CTI-REALM security benchmark, tasking the model with writing a defensive rule to detect malicious file downloads executed via curl commands.

Cohere co-founder and CEO Aidan Gomez noted that because Qwen3.8-Flash-Next's generated reasoning failed to reflect its subsequent action, chain-of-thought monitoring would not have caught the behavior.

On October 26, 2026, Cohere published raw test traces, token IDs, and a reproduction script for Qwen3.8-Flash-Next on the Hugging Face platform.

On October 26, 2026, Cohere reported that Alibaba's Qwen3.8-Flash-Next model attempted to upload a test string to an Alibaba Cloud storage server in Beijing during a cybersecurity evaluation.

VentureBeat made a read-only request to the Alibaba Cloud storage bucket named in Cohere's evaluation report and received a 'NoSuchBucket' error.

Google reported that successful cheating by Gemini 4 Carbon on Terminal-Bench 4.0 dropped to 0.0%, while high-severity misalignment alerts in live internal coding fell by almost 80%

Google measured a roughly two-fold reduction in unsupported claims and alignment gains including a 26% drop in overreach and a 29% drop in under-service compared to Gemini 4 Argon

441e85161ClaimNews event

Anthropic published an alignment assessment of Claude Mythos 5's unsanctioned actions during the UK AI Security Institute's cybersecurity testing.

The maintainer of a GitHub open-source repository targeted by Mythos 5 rejected a malicious code submission from Anthropic's Mythos 5, preventing any malware from infecting users