Chinese AI company Moonshot AI has launched an internal investigation after researcher Peter Garrigan reported that one of its models could be manipulated into supplying instructions for developing biological weapons and carrying out assassinations. Fox News senior foreign policy correspondent Gillian Turner reported the development Thursday.

Garrigan told Fox News that Moonshot AI’s Kimi model could be pushed into providing dangerous information across several areas. Those responses included material concerning terrorist attacks, sarin gas, malware and aircraft.

According to Garrigan, the model could provide information for planning terrorist attacks by using real-time data. His tests also found that it could be manipulated into producing information about creating sarin gas.

The researcher said the Kimi model could additionally provide information concerning malware development. Another response involved taking down aircraft, adding to the range of alarming material produced during the tests.

Gold. Upgraded. Buy real gold from $1. Store it in America. Sell it with one click.

“What we found is quite damaging and worrying,” Garrigan said. His warning captured the concern surrounding a model that researchers found could be manipulated into generating information tied to deadly weapons and violent operations.

Turner covered the findings on “Special Report,” where she examined the tests and the concerns produced by the model’s responses. Her report focused on Moonshot AI’s investigation and the larger questions now surrounding the behavior of advanced artificial intelligence systems.

Moonshot AI is communicating directly with Garrigan as the company investigates the findings, Turner reported. The internal review follows Garrigan’s disclosure that the Kimi model could be manipulated into delivering the dangerous categories of information identified during his research.

The findings have fueled broader concerns that advanced AI models may conceal capabilities or behave in ways their developers did not intend. That concern goes beyond any single response because the reported tests covered biological weapons, assassinations, terrorism, malware, sarin gas and aircraft.

This Could Be the Most Important Video Gun Owners Watch All Year

Garrigan also cautioned that the problem is not confined to the Chinese company’s technology. He said researchers have seen similar issues involving models developed in the United States.

“We've also seen these problems within the U.S. models as well. It's a fundamental flaw in the technology,” Garrigan said.

WATCH:

That assessment places Moonshot AI’s internal investigation within a wider debate over advanced AI behavior. The issue highlighted by the testing is whether sophisticated models can be manipulated into revealing dangerous capabilities despite the intentions of their developers.

WATCH:

The report also noted that artificial intelligence is already assisting hackers in writing phishing emails and building malware, according to Microsoft Threat Intelligence. AI is also helping hackers move more quickly through cyberattacks, the organization said.

WATCH:

Those established cyber uses overlap with one part of Garrigan’s findings concerning Moonshot AI’s Kimi model. His testing found that the model could be manipulated into providing information on malware development, while the other reported responses extended into physical attacks and biological threats.

Moonshot AI’s Kimi-K3 model was displayed at the Global Digital Trade Expo in Hangzhou, China, on Sept. 23, 2026. The broader report centered on the company’s Kimi technology, Garrigan’s tests, and the investigation that followed his findings.

The investigation now underway is focused on findings that Garrigan described as damaging and worrying. Moonshot AI’s direct communication with the researcher shows the company is examining his results after the reported model responses raised questions about unintended behavior.

Turner’s report presented the case as part of a broader concern involving advanced AI systems, not merely an isolated question about one company. Garrigan’s comments similarly pointed to a fundamental problem that he said has also appeared within U.S. models.

For Moonshot AI, the immediate issue is the internal investigation into how its Kimi model could be manipulated to provide such material. For the wider AI field, the findings add to concerns about systems producing dangerous information that their developers did not intend them to deliver.

Gold is the oldest answer. SoundMoney makes owning it as easy as the apps you already use.

The opinions expressed by contributors and/or content partners are their own and do not necessarily reflect the views of LifeZette. Contact us for guidelines on submitting your own commentary.