Last week, a cross-party parliamentary committee published its long-anticipated report on AI regulation. There's plenty in it that in-house B2B marketing teams will want to pay attention to.
Produced by the Joint Committee on Human Rights, Human Rights and the Regulation of AI is the result of a major inquiry involving more than 70 written submissions and 10 oral evidence sessions. It calls for, among other things, a dedicated AI Bill, a single statutory regulator, and mandatory transparency about when and how AI is being used. None of this is law – yet. The government has two months to respond to the report, and the degree to which it will agree with and implement any/all of the recommendations is up for debate.After all, the government has so far preferred regulation at the point of use by existing sector regulators. However, with a new Prime Minister quickly shaking up the government departments related to AI upon his arrival and promoting the AI Minister into a Cabinet role for the first time, it’s hard to tell which way it could go for the report.
That makes the report worth paying attention to so, we’ve been taking some time to get under its skin.
What impact could the AI report have on in-house B2B marketing teams?
Large Language Models (LLMs) and AI platforms and tools have, of course, become a core day-to-day tool for agencies and in-house marketing teams alike, supporting everything from research to content production and creative output. But the parliamentary report argues that the voluntary approach governing some of these applications may need to be underpinned by legislation and increased scrutiny.What exactly that could mean in practice? Well, here’s some of the most relevant aspects we’ve identified for marketers to have on their radar.
1. Disclosing when and how AI has been used
The committee wants mandatory transparency across the AI lifecycle, with an obligation to state when AI systems are being used in situations where that use could significantly affect individuals, groups or communities.This is potentially relevant to number of common marketing functions – automated customer communications, campaign personalisation, chatbots and AI-assisted service decisions for example.
That said, the idea isn’t that every AI-assisted function needs to be labelled, it's about ensuring people know when a system has shaped something that may specifically affect them.
2. Responsibility across the supply chain
The report highlights that AI supply chains are typically complex and opaque, making it difficult to identify the source of risk and who's responsible for it. Legal risks are currently pushed downstream through contracts and standard terms, leaving upstream developers largely shielded. The committee wants to push more of the burden back up towards them.This conclusion actually works in marketers’ favour. As things stand, the Equality Act 2010 doesn't apply between businesses; so, it may not reach the developers supplying AI systems if a tool produces discriminatory or damaging outputs. The deploying organisation (i.e. potentially the in-house marketer’s company) therefore typically carries the majority of the exposure.
The committee's recommendation is that needs to change and that the obligation should sit at every stage of the life cycle. So B2B marketers need to start thinking about their third party relationships and considering how any risk is mitigated and accounted for. Working with media houses, automation and CRM software partners, agencies and even freelancers. Perhaps it's time to start mapping out where your exposures are and considering how you can cover all your B2B bases before it becomes a legal requirement to do so.
3. Human-in-the-loop is not enough
Data protection already restricts decisions based solely on automated processing where there's no meaningful human involvement. But the committee's view is that the mere presence of the human-in-the-loop doesn't necessarily constitute sufficient ‘human involvement’.That’s because even highly trained reviewers may struggle to scrutinise AI output, given the volume of information involved and the fact that people are prone to automation bias.
For marketing teams, this reinforces that sign-off cannot be treated as a box ticking exercise. The person approving an AI-assisted decision needs to understand the system's limitations and levels of accuracy as well as both the authority and the appetite to reject AI output it considers could be negatively received.
Interestingly, it wouldn’t take new legislation to make this recommendation a reality (the Secretary of State for Digital, Culture, Media and Sport already has powers to change the definition of ‘human involvement’ via the new Data (Use and Access) Act 2025) – but this is an area that could come to fruition and expand much more quickly in the next 6-12 months. So better to play safe, than be sorry. Your start point should be to review all AI application across your marketing ecosystem to prioritise areas of potential exposure and focus on these first.
What to do now to get your 'in-house' in order
As mentioned, none of this is law, yet, and it may never become law. However, most of it is good practice regardless. Here’s some useful learnings and actions we’d recommend in-house marketing teams and marketing leaders taking:- Build an AI use register which identifies which tools are doing what and touching whose data
- Run data protection impact assessments for personalisation or automation decisions that touch identifiable individuals using AI workflows
- Review agency and martech contracts, indemnities etc on training data and ask for full disclosure of the underlying models they use and how
- Speak to agency partners about their AI-usage policy, guidelines, processes and training and how they record and ensure this
- Check or agree who holds the audit trail on all relevant AI processes within your B2B marketing team and agree a written disclosure standard now rather than retrofitting one later
- Name your reviewers in your policy, train them in each tool's limitations, ethical decision-making (particularly around data usage) and make clear they should reject output which potentially breaches any AI policy guidelines
It's also important to note that the committee stated that lower risk systems should face fewer requirements – it doesn’t want to add unnecessary admin burdens. Most day-to-day marketing use of AI would likely sit at that lighter end.
Bart Maginn
Growth & Innovation Director
Share this article
Make sure your team is ahead of the game on AI usage best practice
Marketing leaders already know their teams need training and guidance, balanced against the freedom to experiment, when it comes to maximising the potential of AI in safe and secure ways. But this report is a reminder that scrutiny is only going one way and that in-house B2B teams which have clear processes now will have the least to retrofit or justify down the line. We believe regulation and scrutiny around AI usage, LLM output and tools will continue to increase and quickly.If you’d like to talk about how you can ensure your team is equipped to safely and effectively use AI to create marketing advantage, please get in touch hello@uppb2b.co.uk.
