Although this is one state-based law, other US jurisdictions are developing rules to govern AI-altered content that is applicable to any funds generating material.
TL;DR: check for ‘AI provenance’ in compliance approvals.
Recently, Pennsylvania has signed an AI Deceptive Advertising Act, requiring advertisers to label commercial content generated or altered by AI. Specifically, if that content could give a false impression to consumers.
The fact that pretty much any form of synthetic content can be made using AI these days, this feels like a foregone conclusion that will ripple outside of the state. Already, California and New York are developing their own similar rules.
There is no exemption for certain fund sizes, although text-only content, parody and clearly fictional material falls outside of its remit.
What the act stipulates is that labels must be conspicuous, clear, appear at the first point of exposure, be hard to remove, and use the same language and medium as the ad. The Attorney General or district attorney can seek injunctions and civil fines.
So, what does this mean for fund marketers?
Reasoning under existing rules
At the same time, many regulations pertaining to AI use or the clarity of communications are already in force for fund marketers to abide by:
- The SEC Marketing Rule bans statements that are materially misleading, or implying something that is untrue.
- FINRA 2210 requires fair and balanced communications.
- The FCA requires similar, with its Consumer Duty asking firms to test how retail customers actually understand what they read.
- The EU AI Act’s transparency obligations for GenAI content point the same way for firms that distribute in Europe.
Are we moving toward GenAI-only content?
Any AI-generated ‘portfolio managers’ or voiceovers used to present a testimonial or strategy video carries the obvious risk under many regimes. This will be the first thought at many firms before delving into GenAI video content experiments.
But those that are thinking of using synthetic voices, actors, and dynamic charts must be aware that these could be considered misleading, simply because the audience may not necessarily know that a machine made them.
On the practical front, compliance will find it a bit of a head-scratcher to have to approve how an asset was produced rather than just a set of words. Many current review workflows have no field that records if an image, voice or animation is computer generated – though this should certainly change.
Recommendations to stick by
This Pennsylvania rule will not be a problem for a lot of fund managers. Institutional campaigns are not consumer advertising, and the text exemption covers a lot of what asset managers usefully publish.
Still, it is worth considering the following:
- Add an ‘AI provenance’ field to compliance submission forms, in order to track anything generated or edited using AI tools (which should also be disclosed).
- Set a house disclosure/label standard for any synthetic content that gets promoted in retail-facing channels.
- Never AI-generate managers or clients for use in performance or product content.
- Brief agencies and content freelancers with AI disclosures, and make them a contractual requirement.
- Store AI prompts, source files and approval records together to fulfil SEC and FCA record-keeping duties.
Source
Constellation Research, “AND AWAY WE GO: Pennsylvania Signs AI Deceptive Advertising Act. What it Means for Brands” (via TLDR Marketing)
