Sneha Revanur, founder of the AI advocacy nonprofit Encode at age 15, shares her journey from initially focusing on immediate societal harms of AI, like criminal justice issues, to now prioritizing existential risks posed by advanced AI systems. Her shift was catalyzed by firsthand experience with large language models like ChatGPT, which challenged her earlier skepticism about AI capabilities. She emphasizes the importance of adopting a mindset open to extrapolation and imagination to anticipate rapid AI progress and its potential dangers.
Encode’s approach to influencing AI policy balances inside and outside political strategies. While traditional legislative processes are slow, Encode focuses on building power through creative coalition-building and comprehensive involvement—from policy development to implementation. Sneha highlights successes such as helping pass landmark AI safety laws in California and New York, which, beyond their immediate effects, build regulatory capacity in key states that can act swiftly if federal action lags. She acknowledges the challenges of advocacy, including facing well-funded industry lobbying and the frustration of setbacks, but stresses that persistence and learning from failure are crucial.
A key part of Encode’s strategy involves coalition-building across diverse groups with differing interests, such as family-first conservatives, affected parents, and entertainers concerned about AI harms. Sneha explains that while this diversity risks diluting goals, clear internal principles and strong relationships based on trust help navigate conflicts and maintain focus. She also discusses the tradeoffs in legislative advocacy, noting that while some bills like California’s SB-53 were weaker than initial versions, they laid important groundwork for future, stronger laws in other states and at the federal level.
Sneha recounts a tense episode when OpenAI subpoenaed Encode’s general counsel during the SB-53 campaign, which brought significant public attention and underscored the high stakes of AI advocacy. Despite initial intimidation, Encode maintained a balanced stance to preserve future cooperation with OpenAI, which has since shown signs of shifting toward more constructive engagement on regulation. She observes that internal employee pressure and civil society advocacy are influencing companies to reconsider their political strategies, moving away from outright opposition to AI regulation.
Looking ahead, Sneha stresses the growing political salience of AI and Encode’s focus on supporting candidates who address both near-term and existential AI risks. She highlights the importance of preparing all branches of government to handle AI’s challenges to avoid dangerous power concentration. On strategy, she advocates for sequencing policy efforts—starting with achievable wins like whistleblower protections while preparing for more aggressive interventions like liability rules as political will grows. Finally, she encourages new advocates to balance analytical rigor with political persuasion, build knowledge through community engagement, and embrace both “soldier” and “scout” mindsets to effectively influence AI governance.
