Britain’s parliamentary Science, Innovation and Technology Committee has issued a blunt message: the public agencies expected to oversee artificial intelligence do not have the money they need to keep up with the technology’s rapid expansion. The committee chair warned that regulators are being asked to respond to fast-moving systems without the financial backing required to do the job properly.
At the center of the concern is a funding pledge the government announced in February, directing £10 million to support the Office of Communications and other authorities as AI use and capability accelerates. In the committee’s view, that sum does not match the scale of the challenge, and it described the support as clearly insufficient for effective oversight.
The committee also pointed to an emerging strain in the UK’s safety architecture: it said there were reports that some developers’ models were not being made available to the newly formed AI Safety Institute for pre-deployment safety checks. For a regulatory ecosystem that relies on early visibility into advanced systems, the suggestion that certain models are being withheld underscores how fragile the current arrangements may be when cooperation is uneven.
A Funding Gap With Real Consequences
In the report’s assessment, the next administration should be prepared to state plainly what additional funding will be provided, at a level that reflects the magnitude of the work regulators are being asked to perform. The implication is not simply that AI governance is expensive, but that under-resourcing can turn policy into performance, with watchdogs always reacting after the fact.
Alongside the call for more money, the committee raised the question of access, focusing on claims that regulators and the AI Safety Institute may not be seeing certain models before they are deployed. If that access problem persists, the gap will not just be financial; it will be informational, limiting regulators’ ability to assess risks in advance rather than responding once harm is already visible.
The report further said that, because the Bletchley Park conference in November 2023 involved an agreement around this kind of cooperation, the incoming administration should identify which developers are refusing access and require an explanation for the refusal. In other words, the committee framed transparency about non-cooperation as part of the governance task, not a side issue.
Deepfakes, Elections, and the “Black Box” Problem
Beyond budgets and access, the committee’s report highlighted what it described as AI’s capacity for deception, with a particular focus on deepfakes. It warned that manipulated content can be aimed at undermining democratic processes, and it said government and regulators should take firm enforcement action against online platforms that host such material in order to protect the integrity of the general election campaign.
That recommendation reflects a view that the most immediate risks are not always futuristic, but operational, showing up in information ecosystems that shape public trust. The committee’s language suggests it sees enforcement not as an optional extra, but as a necessary response when platforms facilitate the spread of harmful synthetic media.
The report also flagged a deeper technical and governance challenge: AI systems that behave like a “black box,” where the reasoning behind an output may be unclear or unknown. It described this opacity as perhaps the most significant difficulty, because oversight becomes harder when neither regulators nor the public can easily trace how a system arrived at a particular decision or conclusion.
Fresh Warnings From AI Safety Voices
The committee report arrived shortly after another public warning, issued on May 20, about what it described as insufficient safeguards in the event of a major AI breakthrough. That warning came from a group of twenty-five experts, including Geoffrey Hinton and Yoshua Bengio, identified as two of the three “godfathers of AI,” both recipients of the ACM Turing Award.
The scholarly work set out government safety frameworks designed to tighten standards if technology advances quickly and extreme risks grow during periods of rapid capability gains. It also called for stricter risk-checking expectations for tech companies, increased funding for newly established bodies such as AI safety institutes in the US and the UK, and limits on the use of autonomous AI systems in critical societal roles.
Additional co-authors named in the piece include Yuval Noah Harari, Daniel Kahneman, Sheila McIlraith, and Dawn Song. Their warning argued that society is investing heavily in making AI systems more capable while spending far less on making them safe and minimizing harmful impacts, concluding that “we” are not adequately prepared for the dangers.
