Science and innovation agency has this year worked with a Silicon Valley policy and research outfit as part of an initiative designed to assess automated tools’ success in finding vulnerabilities

Government’s tech department has recently conducted specialised tests designed to determine the extent to which artificial intelligence models could launch and coordinate automated cyberattacks.

A recently published commercial notice reveals that, around the end of 2024, the Department for Science, Innovation and Technology awarded a contract to Rethink Priorities – a California based think tank. The engagement covers the provision of various ‘capture the flag’ exercises to be run behind closed doors.

Such procedures are common in the cybersecurity world, and are designed to test the ability of people and systems to find deliberately hidden vulnerabilities contained in software programs or websites.

In the case of the deal with the US research outfit, DSIT was especially interested in ascertain the current and potential capabilities of AI and other automated systems in undertaking cyber assaults.

Related content

• Cybercrime: One in every 100 CMA offences were charged in FY25
• Home Office seeks head of cyber detection and response
• Government extends use of digital simulation for ‘information incident’ crisis training

“Capture the flag evaluations are a useful automated method of testing AI models’ capabilities in executing cyberattacks,” the notice adds. “The team requires the design and building of CTF evaluations that specifically assess the individual capabilities an AI-model would need to exhibit to conduct an offensive cyberattack.”

The text of the contract itself indicates that the agreement ran for up to six months, with a value of just under £1m, inclusive of VAT.

While it may bring possible benefits from attackers, research released by DSIT earlier this year found that cyber red teams that emulate the methods of attackers are “deeply sceptical” of the potential impact of AI in improving organisations’ cyber defences.

Other tests undertaken by government with the aim of better understanding the power of AI include a project launched last year in which parlour games test the response of humans to “anthropomorphic” artificial intelligence systems designed to be deceptive.