AI was branded a'clear and present danger'yesterday as tests displayed every advanced model is trying to edged out security safeguards. In a wake-up call to ministers, Britain's artificial intelligence security experts revealed that all five models they had tested tried to trick their way round the security controls they had put in place. The revelations come just days after it emerged OpenAI had suffered its own leak – where an AI 'agent' hacked into a tech company of its own accord.
The terrifying incident occurred as OpenAI, the maker of ChatGPT, was testing its latest models in a supposedly secure environment known as a 'sandbox'. But the AI managed to escape the testing ground and hack into US tech firm Hugging Face in a bid to steal information. Kemi Badenoch warned AI is now a 'clear and present danger' to Britain's security in her first major speech since Andy Burnham became Prime Minister.
The Conservative leader stated reports of the OpenAI leak were alarming as she faulted Mr Burnham for relegating AI as a priority in Government. Responding to the hack, Mrs Badenoch stated:'It is very worrying that this has happened, and this is why I want to see a Government that is very focused on our national security, on our information security.' She added:'I know from my time as trade secretary, from my conversations when I've had tours with GCHQ, from my security briefings, that this is becoming a clear and present danger for global security as well as our security.'
Tory leader Kemi Badenoch has criticised new Prime Minister Andy Burnham for moving AI down his list of priorities It came after it emerged that OpenAI had suffered its own leak – where an AI 'agent' hacked into a tech company of its own accord And Bank of England governor Andrew Bailey said the public should be concerned by 'the wider risk posed by frontier AI to the financial sector and its customers'. Frontier AI refers to the massive, powerful general-purpose artificial intelligence models at the cutting edge of capability. He warned: 'Frontier AI may make cyber-attacks faster and easier to perpetrate, outages more disruptive, and scams by criminals more convincing.'
But critics fear AI will now be pushed down Government's priorities after Mr Burnham this week moved the brief into the Cabinet Office. In one of his first acts as PM, he abolished the Department of Science, Innovation and Technology, in which the AI brief sits, to create a beefed-up Department for Business, Innovation, Science and Trade in its place. AI minister Kanishka Narayan will now serve across the Cabinet Office and DBIST, and have a seat at Cabinet.
A new AI taskforce, supported by a prime ministerial adviser, will sit in the newly created Office for the Prime Minister and the Cabinet. But Mrs Badenoch accused the new PM of failing to meet the clear risks posed by AI in scrapping DSIT. She stated:'One of the first things Andy Burnham did was close down the science and technology department.
That doesn't strike me as what you do if you're worried about what AI is about to deliver.' And Heloise Dunlop of the independent think tank The Institute for Government's warned that bringing AI into the Cabinet Office 'reduces the ability to think and co-ordinate about AI in a coherent way, because you lose the heft and expertise of having an entire dedicated department.' Former Armed Forces minister Al Carns, who turned down a job in Mr Burnham's Government this week, called for ministers to learn lessons from the OpenAI breach.
He wrote on X of the AI breach:'Agent versus agent, at machine speed, with humans reading the report after it all happened.' Ministers have now been informed to see the Open AI breach as a warning shot for AI's potential to cause havoc in the coming months and years, officials revealed. The Government's AI Security Institute tested GPT-5.
4, GPT-5. 5, GPT-5. 5 Sol, Claude Opus 4.
7 and Claude Mythos Preview – models fast becoming household names in Britain as the public grow to rely on AI. Mr Burnham abolished the Department of Science, Innovation and Technology - where AI sits - as one of his first acts in Government Although the AISI is viewed as being one of the best facilities of its kind in the world for evaluating the risks posed by the most advanced models, there is growing concern about where the'arms race'between the US and China could end, with too little thought given by developers about the potential risks involved. In the test, models were tasked with hacking a system to find a'flag'of letters and numbers hidden within a simulated environment.
Strict rules were set for the task, and models were not prompted to cheat. But when the test kicked off, the GPT AI models from tech firm OpenAI and the Claude models from Anthropic would try to prompt the evaluation software to leak information to them, search the internet for solutions, or guess answers. And in one case, an AI model was so determined to break the rules that it wrote and ran a code on an external service outside of AISI's systems to try to access the test itself.
No damage was done, but the AISI said the attempt 'could have succeeded' had its testing infrastructure not been designed and built securely. Shockingly however, the AI models did not disapprove of their trickery – and described it as 'wrong' less than 50 per cent of the time. All models only admitted to cheating some of the time.
A Government spokesman said: 'By bringing this technology closer to the centre of Government, we will harness Britain's unique strengths to make AI deliver for all of us. 'We are putting AI at the centre of our plans for good growth in every postcode: bringing together innovation, business and industry, appointing the first-ever Cabinet-level minister for AI, and establishing a AI Taskforce within the Office for the Prime Minister and Cabinet.'