\n To print this article, all you need is to be registered or login on Mondaq.com.
\n <\/p>\n
Article Insights<\/p>\n
Lee Ramsay\u2019s articles from Lewis Silkin are most popular:<\/p>\n
\n with Senior Company Executives, HR and Finance and Tax Executives
\n in United Kingdom
\n with readers working within the Healthcare, Property and Securities & Investment industries
<\/p>\n
Somewhere on a leafy British road the above sign sits in quiet
\nabsurdity. Drivers read it, ponder its existence, and accelerate
\naway none the wiser. It’s a sign about a sign that isn’t in
\nuse.<\/p>\n
But all is not as strange as it first seems. Typically, the sign
\nis a placeholder which indicates that the variable message sign on
\na nearby gantry isn’t working yet or is being tested.<\/p>\n
AI regulation in the UK finds itself in similar territory: there
Put simply, the UK is still in testing mode on AI regulation,
This is why the\u00a0Digital Regulation Cooperation
What follows is our second article on the practical lessons for
What is the DRCF?<\/p>\n The DRCF launched in 2020 as a voluntary forum, to ensure a
Agentic AI<\/p>\n With the recent publication of the\u00a0ICO’s Tech Futures report on Agentic
Accountability in the agentic AI supply chain is one senior
Place agentic AI in the consumer context and then you
While agentic AI presents opportunities, senior leaders need to
It is imperative to protect consumers from harm, whether that
If businesses don’t get this right they will likely find
Consumer trust is paramount and at the moment trustworthiness in
Children’s wellbeing\u00a0<\/p>\n With the backdrop of the government’s consultation on UK
Depending on your point of view there were some
Children exhibit a high level of trust in chatbots, often
This is why the ICO’s\u00a0Age Appropriate Design Code<\/a>\u00a0is so
Well-designed chatbots, that don’t process children’s
The discussion concluded that children and parents need to
UK Government AI roadmap<\/p>\n
\nare signs that things are afoot, but concrete signposting seems
\nsome way off. The\u00a0Financial Times<\/a>\u00a0recently reported that an
\nAI Bill isn’t looking likely in the next King’s Speech,
\nwhich is due to take place in mid-May.\u00a0<\/p>\n
\ntrying to work out where it wants to go and how it wants to get
\nthere.<\/p>\n
\nForum<\/a>\u00a0(DRCF) has assumed such importance in shaping
\nBritain’s current approach to AI. On 10 March 2026, the forum
\nconvened its second Responsible AI Forum in which all those
\ninvolved in AI could better understand the who, what, when, where,
\nwhy and how of UK AI rules in 2026.<\/p>\n
\nsenior leadership teams and general counsel who need to understand
\n(and thrive in) the shifting world of AI regulation.<\/p>\n
\ngreater level of cooperation between regulators, given the unique
\nchallenges posed by regulation of online platforms. It brings
\ntogether four UK regulators with responsibilities for digital
\noversight: the Competition and Markets Authority; the Financial
\nConduct Authority; the Information Commissioner’s Office; and
\nOfcom.<\/p>\n
\nAI<\/a>\u00a0we have some insight into the regulatory thinking
\naround this emerging technology. While the ICO work focusses on the
\ndata protection implications of deploying AI it was interesting to
\nhear from Ofcom, UKAI and Responsible Intelligence about trends and
\ndevelopments of agentic AI.<\/p>\n
\nleaders will be pondering before embarking on any projects. It is
\nclear there needs to be discussion around where boundaries lie
\nwithin the agentic ecosystem and the web of relationships this
\nemerging tech inevitably brings. If something were to go wrong
\n\u2013 and there are already many examples to choose from \u2013
\nhow will liability be assessed and determined and how will this
\nreflect on your brand’s reputation as well as the bottom
\nline?<\/p>\n
\nsignificantly up the ante. Transparency and explainability will be
\nkey to compliance but also so consumers can understand exactly what
\nit is they are signing up to and how they will be protected in line
\nwith existing laws and guidance. We were directed to several
\npublications on agentic AI recently released by the CMA which
\nprovide insight into the regulator’s thinking:<\/p>\n
\nstep back and take time to make informed decisions. It is important
\nto experiment, but also build in pilot phases and extensive testing
\nand feedback loops before launching such a product.\u00a0<\/p>\n
\nharm is caused directly or indirectly, to avoid potential
\nexploitation, particularly for vulnerable consumers who may not
\nunderstand the possible outcome of using agentic AI, as well as
\nensuring competition is across the whole AI stack so consumers are
\nable to give clear consent to transparent decision making with
\ntheir best interests at heart.\u00a0<\/p>\n
\nthemselves falling foul of existing laws and potential accusations
\nof manipulation, price discrimination and lack of competition in
\nthe goods and services offered.<\/p>\n
\nAI, let alone agentic AI, is a huge issue. Many people are aware of
\nfraud and misinformation but when it comes to agentic AI, unless
\nthere is transparency and explainability, as well as the right to
\nchallenge decisions, it seems the uptake will not match the hype.
\nThat said, responsible development with implemented safeguards,
\nallowing for genuine, freely given, informed consent might see a
\nvery different landscape in 12 months’ time.\u00a0<\/p>\n
\nchildren’s digital wellbeing, covering social media age bans,
\ncurfews, AI chatbots and gaming (for more see our article\u00a0here<\/a>), the DRCF hosted a session on growing up
\nwith chatbots.\u00a0<\/p>\n
\nfascinating\/terrifying statistics shared by the expert panel, e.g.
\none third of UK teenagers use a chatbot for an emotional
\nrelationship, 56% of UK teenagers believe AI can think, 23% believe
\nAI can feel emotions and 40% have no concerns about taking advice
\nfrom an AI chatbot.<\/p>\n
\nblurring the boundary between what is a chatbot and what is a
\nfriend. While LLMs are improving all the time, they don’t have
\nempathy \u2013 they are merely learning how to respond to
\nemotional questions, and we know they don’t always get the
\nresponse right.<\/p>\n
\nimportant and for those exploring AI tools that are in scope, it is
\nessential to invest time and resources in getting this right. Media
\nand regulatory scrutiny in this area are at an all time high
\nglobally and no-one wants to be making headlines for the wrong
\nreasons.\u00a0<\/p>\n
\nspecial category data, have safety by design features enabled, and
\nare segmented to deal with cognitive development do have a positive
\nrole to play in children’s lives.<\/p>\n
\nupskill to understand the tech the children are being exposed to
\nand have informed conversations around its use. Companies operating
\nin this space need to carefully consider their legal, regulatory
\nand ethical obligations and ensure they always put the best
\ninterests of the child first. This raises nuanced questions about
\nrevenue models, creating dependency, doom scrolls etc. but that, as
\nthey say, is for another day!\u00a0<\/p>\n