The Results are in: AI on the Front Line of Alcohol Advertising Regulation
The UK’s Advertising Standards Authority (ASA) has recently published the findings from its latest trial, which utilized artificial intelligence (AI) to evaluate the compliance of alcohol-related advertisements with established regulations. This trial represents the ASA’s most extensive deployment of AI-assisted monitoring to date, highlighting its commitment to innovation as AI is now regarded as a core part of the organization’s regulatory toolkit.
By employing large language models, referred to as the AI-agent, to analyze the content of advertisements against a comprehensive section of the CAP code, the ASA effectively identified potential breaches before any public complaints were lodged. This proactive approach allows the ASA to remain agile and responsive as a regulator in a landscape where AI-generated content is increasingly prevalent.
The ASA Trial: Key Findings
The AI-agent assessed the text and imagery of nearly 6,000 online paid-for UK ads in a matter of minutes, specifically analyzing them against section 18 of the CAP code. To facilitate accurate assessments, the AI-agent was provided with contextual guidance, including illustrative examples derived from previous ASA rulings on alcohol.
Key findings from the trial include:
- High Compliance Rate: Overall compliance for alcohol advertising was notably high, with 96% of the reviewed ads deemed likely to comply with the CAP Code.
- Alcohol-Free Alternatives: In contrast, only 48% of the reviewed alcohol-free ads were found to have potential compliance issues.
- False Positives: The AI-agent flagged 40% of ads as concerning, which were later determined to contain possible breaches upon expert review.
- False Negatives: The AI-agent successfully flagged 65% of content with potential issues, indicating a significant recall rate.
Notably, the AI-agent identified non-compliant ads that had not previously generated any public complaints, showcasing its extensive reach. The most common issue among alcohol-free ads was unclear or missing ABV information, as outlined in Rule 18.19 of the Code, which mandates that marketing communications for alcohol alternatives must include a clear statement of their ABV levels.
Looking Ahead
The ASA’s integration of AI-assisted regulation is likely to persist. The trial illustrates that AI can complement human expertise in assessing compliance within specific advertising sectors at scale. However, this remains a complex endeavor, balancing the advantages of AI with the necessity for continuous testing and validation to address inaccuracies.
Providers within the alcohol advertisement sector should anticipate the following developments:
- Increased Sector-Specific Guidance: The ASA’s ability to efficiently identify problematic themes is expected to lead to further targeted guidance. An article has already been released addressing common breaches identified for no/low alcoholic drink content, clarifying distinct regulatory requirements within rules 18.18 – 18.24 of the CAP Code.
- Shift to Continuous Compliance: Brands should operate under the assumption that their advertising is under continuous scrutiny, with the ASA now capable of conducting “market sweeps” across sensitive categories like the alcohol industry.
- Shorter Publication-to-Enforcement Cycles: As the ASA’s reach expands and compliance issues are flagged prior to receiving complaints, advertisers must familiarize themselves with the relevant CAP code, knowing that this is the content the AI-agent is trained on.
