Several members of our team who work on content trust and safety recently traveled to San Francisco for the fifth annual TrustCon, presented by the Trust and Safety Professional Association.

TrustCon is the largest gathering of trust and safety professionals in the industry that brings together online platforms, local and international regulators, and specialized vendors to discuss solutions necessary to keep online spaces safe. Members of our team also took to the stage to share some of our own work in this space with the trust and safety community. 

After three days and nearly 180 sessions, our team returned with a clearer sense of where trust and safety is heading for us at Scribd, Inc., and the broader industry as a whole. Here’s what stood out to us:

• AI models are only as good as the policy - and the people - behind them. The real bottleneck has shifted from "can the model apply the rule" to "is the rule written clearly enough to apply." And even the best models have blind spots, so teams are increasingly comparing multiple models' judgments rather than relying on one. Disagreement between those judgments, human or AI, is a useful signal, not just noise - often it's where policy is unclear or a system needs retraining.
• More companies are investing in smarter classification of content by risk level, directing deeper review toward the material most likely to need it rather than treating everything with a uniform level of attention and urgency.
• Trust and Safety isn’t something you bolt on after growth - the cost of getting it wrong is the loss of trust itself. Companies are generally shifting from “firefighting” toward “fire prevention,” positioning trust and safety initiatives as long-term, strategic investments rather than costly afterthoughts. 
• Regulation is accelerating and fragmenting. Age verification, child safety rules, and platform accountability laws are developing rapidly around the world in varying directions. It's a challenge global online service providers, like Scribd, Inc., need to keep an eye on. 
• Human reviewers aren’t going away, but their roles are changing. As AI tools assume routine content review at high volume, human involvement is shifting from case-by-case decision making toward specialized judgement, quality oversight, and teaching AI systems what “accurate detection” looks like.
• There is no "one-size-fits-all" response to content violations. Service providers are building a wider, more nuanced range of enforcement responses (like limiting visibility or reach) rather than a blunt, binary “remove or keep” decision.

More than any single takeaway, the real value of TrustCon is the rare ability to meet and compare notes with our counterparts at companies of all sizes who are genuinely committed, as we are, to the mission of protecting online communities.

Conversations like these help us stay close to where the industry and its regulators are headed, provide valuable intel for our internal processes, and show up with our trust and safety partners.

We'll keep bringing what we learn back to Scribd, Inc., in our continued efforts toward more robust online trust and safety for our millions of users around the world.