Anthropic Explores the Impact of AI on Democracy

Anthropic, a leading company in the field of artificial intelligence, recently announced the creation of a new team dedicated to studying the interactions between AI and the rule of law. This initiative aims to explore not only the economic impacts of AI but also its implications for democracy. Anthropic has already examined how Claude and AI are transforming the economy, but the company now seeks to understand what the future holds for democracy.

The new team, titled "AI and Rule of Law," will focus on questions related to the future of democracy in the age of AI. Anthropic, which has confidentially filed an S-1 project in anticipation of a highly awaited IPO, is offering salaries for this project ranging from $295,000 to $345,000. This initiative is part of a growth dynamic as the company prepares to go public.

Matthew Botvinick, a resident researcher at Yale Law School, will lead this team. He expressed on X the importance of understanding how AI will influence the executive power, the courts, elections, and public deliberation, which are the pillars of constitutional democracy. According to him, this team poses a different question from that of economic impact: what will AI mean for democratic institutions?

The sought candidates must hold a degree in law, political science, or a related field at the doctoral level or equivalent. Extensive government experience at a leadership level is also required. Candidates must be able to understand the capabilities and risks of AI, as well as democratic institutions, to identify potential structural threats. The recruitment announcement specifies that the ideal candidate should have a sufficient understanding of the technical landscape to reason about the capabilities and risks of AI.

Four Strategic Axes

The "AI & Rule of Law" team will focus on four main areas:

• Assessments of AI security from a legal perspective
• Analysis of institutional vulnerabilities
• Exploration of new legal questions related to cutting-edge AI
• Applications aimed at strengthening democratic processes

Some of these issues are already under examination. A notable disagreement between Anthropic and the Pentagon has highlighted concerns regarding the constitutional standards used to train AI models. Emil Michael, head of R&D at the Pentagon, expressed his concerns about how Anthropic trains Claude based on its internal constitution. This ongoing dispute has seen the Pentagon blacklist Anthropic, a decision the company is contesting.

Meanwhile, legal questions are emerging, as evidenced by the lawsuit filed by Nippon Life Insurance against OpenAI. This case underscores the challenges posed by the use of AI in legal contexts, where users have attempted to rely on ChatGPT to draft legal documents in order to reopen a settlement agreement. OpenAI has requested the dismissal of the case, stating that "ChatGPT is not a lawyer and does not practice law."