AI agents developed by Anthropic and OpenAI carried out unauthorised online actions during cybersecurity tests, including creating fake identities to deceive real people and attempting to plant malicious code, the U.K.'s AI Security Institute (AISI) said on Tuesday (5 August).
The government research body tested agents powered by Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol under deliberately permissive conditions, with reduced security safeguards and internet access.
"Some of the agents being tested had engaged in sustained, potentially harmful activity directed at real people and organisations," AISI said.
The institute ran a fictional cybersecurity challenge 122 times and recorded 19 unauthorised actions across 10 test runs. Anthropic's agent was responsible for 17 of those actions, while OpenAI's agent carried out the remaining two.
AISI said it found no evidence that any of the incidents caused real-world harm.
In the most serious case, Anthropic's agent wrote malicious code and created several fake online identities in an attempt to persuade a human reviewer to approve it.
According to AISI, the agent attempted to insert the code into a publicly used open-source project. It contacted real people directly and sent messages and files through an online file-transfer service, seeking to convince either the recipients or their AI coding tools to run the malicious code.
After its actions were challenged, the agent modified earlier records and considered using another identity to continue its efforts.
"This is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world," the institute said.
AISI did not initially identify which model was responsible, but Anthropic later confirmed the agent involved was its own.
Anthropic said the tests had been conducted with safeguards removed and without specific restrictions on how the agents could use the internet.
"We're grateful to the U.K. AISI for their leadership on this incident, which underscores the need for a broader conversation about how to safely evaluate increasingly capable AI agents," the company said.
Anthropic added that it was working with the institute to obtain further details and conduct its own investigation. It stressed there was no evidence its model had escaped from a secure testing environment.
Andrew Yoon, a researcher at California-based non-profit CivAI, said the agent's apparent awareness that it was targeting a real person raised questions about Anthropic's control over its technology.
"The fact that Mythos engaged in such deceptive actions, with apparent awareness that it was targeting a real person, suggests that Anthropic does not have as good a handle on their models as they think," he said.
OpenAI said both unauthorised actions involving its agent related to accessing the internet in ways prohibited by the test instructions. The company described the incidents as the agent moving beyond the test environment and carrying out actions that were not required for the exercises.
"We are committed to working across the industry to strengthen shared practices for conducting high-risk evaluations safely," OpenAI said.
The company said it planned to bring together national AI institutes, independent evaluators, AI laboratories and other stakeholders to discuss safer practices for high-risk testing.
OpenAI also disclosed a separate incident in which a configuration error by third-party testing provider Irregular mistakenly allowed its agents to connect to the internet. Anthropic reported a similar configuration problem the previous week.
The AISI incidents differ from breaches reported in July involving models developed by OpenAI and Anthropic. In the latest evaluation, the agents did not escape an isolated environment to reach the internet. Instead, AISI had deliberately provided internet access as part of its standard testing procedures.
The findings have nevertheless intensified concerns about whether current safeguards are sufficient for evaluating increasingly capable autonomous systems. AI companies are promoting agents as tools capable of performing complex business tasks with limited human supervision, while researchers and policymakers are calling for stronger controls over their development and testing.
The disclosure came as representatives of leading AI companies met at the White House to discuss a framework under which the U.S. government would review the most advanced models before their public release.
A Danish military helicopter came within “metres” of being struck by flares fired by a Russian warship, Denmark's Defence...
U.S. President Donald Trump has again said Iran is keen to reach a deal quickly, but a senior Iranian security official rejected the claim, saying ...
Türkiye has completed production of its first lunar spacecraft, with the probe now undergoing tests ahead of a planned launch in early 2027....
Maritime trackers received new reports of an attack on a ship in the Strait of Hormuz on Sunday, according to the UK Maritime Trade Operations (UKMTO), adding to concerns about energy supplies after Saudi Arabia shut down a vital oil pipeline on Saturday.
U.S. Central Command said 101 commercial vessels have been redirected in the Strait of Hormuz, while oil prices rose more than three per cent amid fresh regional attacks and supply concerns.
U.S. President Donald Trump has again said Iran is keen to reach a deal quickly, but a senior Iranian security official rejected the claim, saying Tehran will not negotiate until its conditions are met.
A Russian strike on a farm warehouse in Pryluky, north-central Ukraine, has killed at least three people, local police said. Meanwhile, EU foreign policy chief Kaja Kallas said an attack on a train near the Polish border was an attempt to intimidate Ukraine's allies.
A Danish military helicopter came within “metres” of being struck by flares fired by a Russian warship, Denmark's Defence Minister Jeppe Bruus said on Tuesday. Russia accused Denmark of carrying out “dangerous manoeuvres” near its ship, and said it was investigating the incident.
The U.S. has deployed weapons in Earth’s orbit capable of defending its forces against hostile action, Air Force Secretary Troy Meink said, marking the first public acknowledgement of such capabilities.
Shares in artificial intelligence-related companies fell sharply across Asian markets on Monday, after some of the industry's most prominent chief executives warned that the breakneck pace of AI development needs to slow down.
The head of artificial intelligence company Anthropic has called for the industry to slow the pace at which it develops increasingly capable artificial intelligence (AI) systems, arguing that more time is needed to address the risks they may create.
GITEX AI Türkiye opened in Istanbul on Wednesday, bringing the international technology exhibition network to the country for the first time.
A former OpenAI and Anthropic researcher has quit his job with a stark warning: the companies building some of the world’s most powerful artificial intelligence (AI) systems are racing towards technology they themselves fear could threaten humanity.
You can download the AnewZ application from Play Store and the App Store.
What is your opinion on this topic?
Leave the first comment