AI Security Competition: Tech Giants Seek Self-Regulation

Microsoft, OpenAI, Google, and Anthropic have recently appointed Chris Meserole as the executive director of their new industry alliance, the "Frontiers of Modeling Forum," according to a recent report by the Financial Times. The forum, established by these four tech giants, aims to fill the global regulatory "gap" in artificial intelligence.

In recent years, artificial intelligence has developed rapidly, with systems like ChatGPT demonstrating abilities such as fluent conversation and generating original text. However, the increasing complexity of AI has raised concerns about potential dangers.

Advanced AI systems use large language models to process massive datasets and generate detailed outputs. However, the complexity of these models makes their internal operations opaque, leading to unexpected behavioral risks.

Meserole told the Financial Times, "It will take some time before actual regulatory measures are put in place." He is currently resigning from his position as the head of AI at a Washington think tank. "During this period, we hope to ensure that these systems are built as safely as possible."

The forum will focus on the risks of artificial intelligence, such as AI's potential contribution to designing bioweapons or generating malicious computer code, which could lead to hacker attacks.

The alliance has committed to investing $10 million in an AI safety fund, with support from former Google CEO Eric Schmidt. The fund will finance academic research on technologies like "red team testing," a method for systematically testing the flaws, vulnerabilities, and dangers of AI systems before their actual deployment.

In addition to funding research, the forum also plans to establish its own technical risk assessment standards and AI system evaluation criteria. Meserole stated that as governments around the world begin to regulate this technology, establishing global standards will provide consistency across jurisdictions. The forum's work aims to complement future government regulations.

This forum's announcement comes just before the Global AI Safety Summit next week. With the EU's comprehensive AI legislation expected to be completed as early as next year and governments worldwide calling for legislation, tech companies are striving to take the lead in establishing ethical guidelines and safety practices for AI. The goal is to shape the upcoming regulatory environment and avoid potential restrictive policies.