Agenda

"As Generative AI transitions from production pipelines into core gameplay loops, developers face a paradigm shift in player safety. Integrating live Large Language Models (LLMs) moves the challenge beyond traditional content moderation into the realm of adversarial creativity. When players interact with unscripted AI, both the player and the AI actively probe for weaknesses, testing boundaries in ways static safety systems can not predict.
This session explores the unique safety issues encountered when the ""content"" is also a ""content creator"". It examines how real-time, high-scale LLM interactions differ from standard User-Generated Content (UGC) and highlights the designer's responsibility to protect players from the AI itself. Attendees will gain insights into player behavior patterns at scale and leave with a robust mental model to help them balance player creative freedom with evolving safety concerns in the era of generative play."

Takeaway

Attendees will get a behind the scenes look at some of the safety concerns in using live AI in games. Both from the perspective of players manipulating the AI, and also the perspective of how the AI can manipulate the players.

Intended Audience

Any developers interested in adding elements of live generative AI into their games or anyone curious about the safety implications of gen AI in gaming will benefit from this presentation. No special knowledge of generative AI is required.