Elith Inc. is pleased to announce that a session planned by its Representative Director, CEO & CTO Koki Inoue and Machine Learning Engineer / Researcher Shosei Nakamura—titled "Does Adding More Rules Actually Make LLMs Safer? A Practitioner Discussion on Prompt Intervention"—has been accepted as a Community Activity at "TrustCon 2026," one of the world's largest international conferences in the Trust & Safety field.
TrustCon is an international conference that brings together practitioners and researchers from around the world who work to improve the safety of generative AI and online services.
In recent years, many organizations have added rules and constraints to system prompts as a safety measure for generative AI.
However, adding more rules does not necessarily lead to improved safety, and discussion continues regarding their effectiveness and limitations.
Starting from the question “Does continuing to add rules to AI actually improve safety?”, this session brings together Trust & Safety practitioners and AI safety researchers from around the world to exchange views on the realities and challenges of AI safety measures, including prompt intervention.
Through dialogue among participants, we plan to deepen discussion on the practical challenges and future direction of improving generative AI safety.
Accepted session overview
- Speakers:
- Koki Inoue, Representative Director, CEO & CTO Shosei Nakamura, Machine Learning Engineer / Researcher
- Title:
- Does Adding More Rules Actually Make LLMs Safer? A Practitioner Discussion on Prompt Intervention
- Acceptance format: Community Activity (TSPA Member-led Meetup)
- Event: TrustCon 2026
- Dates: July 20–22, 2026
- Location: San Francisco, USA
- Organizer: Trust & Safety Professional Association (TSPA)
- Official event website: https://www.tspa.org/event/trustcon-2026/
*This Community Activity is conducted as an individual activity, and TrustCon and the Trust & Safety Professional Association (TSPA) do not endorse or support any specific company, product, or service through it.