Operating at the Speed of AI: How Platforms Can Keep Up

Legal Risk Is Only One Kind of Risk

Panelists started with user-facing uses of AI that aren’t illegal but still cause problems. Think of AI-generated product listings on marketplaces that may be misleading, AI-written reviews, IP concerns, and low-quality AI content on social media that drags down the user experience. Volume and scale make every one of these harder to handle.

Furthermore, panelists talked about ways in which AI is making enforcement harder. For example, AI-built websites may use technology targeting to display innocuous content on desktops, which is what most enforcement teams use, while showing problematic content on mobile devices.

Panelists also stressed that legal exposure is just one factor. Reputational risk, user risk, and local cultural standards all shape what a platform should allow, even when no law forces the issue.

AI Is Making Products Harder to See Clearly

One theme came up more than once: AI-generated images make it harder to know what a product actually is.

Panelists described sellers using AI to alter labels and product shapes so listings look compliant and slip past content moderation. In IP-related cases, AI can be used to hide logos. One panelist also noted that marketing images can look completely different from the real product, which makes it hard for reviewers to know what’s actually being sold.

Some AI-manipulated content may not be malicious, so it’s important to have clear policy lines about when content is generally harmless and when it is intended to deceive.

Writing Policy That Keeps Up

The panel also covered how to build policy that adapts as the technology changes, including what early warning signals teams watch for and how global companies can seamlessly work across jurisdictions while still honoring local nuances.

One panelist discussed the importance of creating a “golden rule” on AI policy that sets a global baseline when jurisdictions disagree, but then also allowing flexibility for variances in jurisdictional laws and cultural standards.

AI as Enforcement

Platforms are leaning on AI to enforce their rules, since the scale of the problem is too big for manual review alone. Panelists agreed that public-facing policies aren’t enough to train an AI system. There’s often a wide gap between regulatory language and what a model can act on, so teams have to turn policy into clear, specific criteria that AI tools can learn from.

Broad rules make this harder. One panelist explained that the broader the rule, the more false positives you get, so teams need to teach the model what their instructions and examples actually mean. Zombie knives were raised as an example of a category where broad regulatory language is challenging to put into practice because it can be difficult to differentiate them from other kinds of permissible knives.

Panelists also made the case for keeping people in the loop. One said human review is necessary when building AI agents, as a basic sniff test on what the system produces. Another pointed out that AI’s role goes beyond enforcement to routing content to the right place; it can direct content to the appropriate team with the appropriate level of urgency so that human reviewers can enforce more efficiently.

Get Ahead of the Gray Zone

Gray areas don’t wait for regulators, and your policies shouldn’t either. LegitScript's risk and policy advisory helps platforms build standards they can defend, spot emerging AI risks early, and turn policy into criteria that people and AI tools can both apply. If your team is working through these questions, schedule some time to talk with us.