Behavioral AI Lab launches UN-linked AI safety initiative targeting scams

Behavioral AI Lab launched a new international AI safety initiative in San Francisco on Aug. 5, 2026, focused on detecting AI-enabled manipulation, scams and deception. The effort is listed in the United Nations Global Dialogue on AI Governance and aims to give developers, regulators and Trust & Safety teams practical tools to prevent harm before it happens.

Why it matters: - AI-enabled scams are getting harder to spot as systems become better at persuasion, impersonation and deception. - Behavioral AI Lab is trying to shift AI safety from reactive moderation to prevention based on human behavior and manipulation patterns. - The initiative could affect AI developers, governments, financial institutions, consumer protection agencies and Trust & Safety teams.

What happened: - Behavioral AI Lab launched the Global Partnership for Detecting and Preventing AI-Enabled Manipulation and Scams on Aug. 5, 2026. - The initiative is listed in the AI Dialogue Partnerships Hub, part of the United Nations Global Dialogue on AI Governance. - Behavioral AI Lab serves as the Implementing Entity for the partnership. - The partnership’s first research focus is AI-enabled manipulation and scams, including cryptocurrency investment fraud, romance scams and impersonation attacks. - The broader mission is to advance AI safety through human-centered evaluation and Safety by Design.

The details: - The partnership builds behavioral evaluation methods that look for psychological and behavioral patterns behind manipulation instead of relying mainly on keyword filters. - The Lab says the approach is meant to help AI developers, Trust & Safety professionals, governments and technology companies stop harm before it occurs. - FBI Internet Crime Complaint Center data showed losses related to cryptocurrency fraud exceeded USD 11 billion in 2025, including USD 7.2 billion from cryptocurrency investment fraud. - Behavioral AI Lab says the initiative builds on research that analyzed more than 15,000 scammer messages from more than 150 documented cryptocurrency romance scam cases across multiple languages and cultural contexts. - The research argues that focusing on persuasion and manipulation patterns offers a more robust way to detect adaptive AI-enabled deception. - The partnership will produce open resources for researchers, developers, Trust & Safety teams, financial institutions, consumer protection agencies and policymakers. - Planned outputs include a Behavioral Manipulation Taxonomy v1, multilingual behavioral-risk datasets, behavioral AI evaluation benchmarks, and a Safety by Design framework for AI-enabled manipulation prevention. - The initiative also plans capacity-building workshops in at least three countries, pilot implementations with global partner organizations, and a knowledge-sharing network with at least ten organizations. - Participating organizations include Shisa.AI, a Tokyo-based company focused on multilingual small language models. - Shisa.AI will contribute multilingual small language model development, safety-focused post-training methods, dataset development and automated model evaluation. - Additional research and implementation partners are expected to join as the initiative expands.

Between the lines: - The launch reflects a broader push to treat AI safety as a behavioral problem, not just a content-moderation problem. - The partnership’s focus on multilingual scams suggests the Lab is aiming at threats that cross borders, languages and platforms. - By tying its work to the United Nations governance ecosystem, Behavioral AI Lab is signaling an ambition to influence global AI policy and standards.

What's next: - The partnership will move from launch to building datasets, benchmarks and implementation guidance. - Behavioral AI Lab plans to expand the network with more research and implementation partners. - Pilot programs and workshops are expected to begin across multiple countries as the initiative scales. - More open resources for practitioners and policymakers are expected as the project develops.

The bottom line: - Behavioral AI Lab is betting that AI safety will need behavioral science to keep up with scams, manipulation and deception that traditional tools miss. - The initiative’s UN-linked positioning could help turn that idea into a broader international framework for prevention.