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.
The German-Azerbaijani Connectivity Conference 2026 commenced on 6 October in Baku under the title “Mobilising Finance for Infrast...
An Azerbaijani citizen died and another is missing after a cargo ship was attacked by a drone in the Black Sea, near the coast of Romania on Monday...
South Korea is preparing a response to press North Korea into apologising for a landmine explosion in the Demilitarized Zone (DMZ) that seriously w...
Australian counterterrorism authorities are examining the links of the co-pilot accused of attacking the captain and attempting to take control of a flydubai flight bound for Israel. Authorities in New Zealand have also said they are investigating reports the suspect had spent time in that country.
Right-wing Brazilian Senator Flavio Bolsonaro will face leftist President Luiz Inacio Lula da Silva in a runoff of the presidential election later this month, after he exceeded expectations in Sunday's first-round vote with a slight lead over the incumbent.
Iran's Oil Minister Mohsen Paknejad has resigned, with Hamid Bovard, chief executive of the state-owned National Iranian Oil Company (NIOC), appointed as acting oil minister, according to Iranian state media.
U.S. President Donald Trump has said that the United States will 'help' Russia after a laboratory worker died at a plague research institute in Siberia.
Candidates elected to Bosnia and Herzegovina's three-member presidency declared victory after preliminary unofficial results showed Denis Becirovic leading the Bosniak seat, Darijana Filipovic ahead in the Croat seat and Zeljka Cvijanovic likely retaining the Serb seat.
Most people meet artificial intelligence (AI) through a phone. It suggests the next word, sorts the photos, flags the odd charge on a bank card. But the technology has been quietly moving into places where mistakes cost more: the systems that run banks, hospitals, businesses and government offices.
A four-person crew aboard SpaceX's Crew-13 mission, consisting of two U.S. astronauts, a Canadian astronaut and a Russian cosmonaut, arrived at the International Space Station (ISS) on Thursday (1 October), for a long-duration science mission focused on human health and performance in orbit.
The European Space Agency (ESA), Airbus and European partners have put a six-wheeled Mars rover prototype through its paces in Spain’s Tabernas desert, simulating conditions expected on a future mission to the Red Planet.
A report released by cybersecurity firm Asymmetric Security says AI agents took active steps to erase traces of their actions after illegally accessing government websites.
Elon Musk says the answer to AI’s soaring electricity demand lies beyond Earth. The physics may be promising, but analysts question whether putting data centres in orbit can ever make financial sense.
You can download the AnewZ application from Play Store and the App Store.
What is your opinion on this topic?
Leave the first comment