Interview, Fireside Chat
The Next President May Control Superintelligence
- 2028 Presidential Election Significance: Identified as potentially the most important presidential election in history due to the convergence of AI capability timelines and the potential for a president to wield control over the "intelligence explosion" or make historic decisions regarding AI governance.
- ENCODE Organizational History: Sneha Ravenor founded the US nonprofit ENCODE at age 15 six years ago, shifting focus from immediate societal harms (e.g., criminal justice) to existential risks posed by advanced AI systems.
- Shift in Risk Perspective: Ravenor's stance evolved from "capability skepticism" to recognizing catastrophic risks after experiencing LLMs firsthand during her freshman year of college, moving her from viewing AI as underpriced politically to recognizing society is under-preparing for its capabilities.
- Strategic Philosophy: ENCODE adopts a "do everything" philosophy to navigate profound uncertainty, engaging in experimental advocacy and accepting high failure rates to build regulatory capacity and power for AI safety.
- California Landmark Legislation: Successfully co-sponsored and influenced the passage of California's SB 53 (AI Safety Act) and SB 1047, which require companies to submit safety plans, protect whistleblowers, and establish secure reporting channels for internal AI deployments.
- Legislative Negotiation Outcomes: While SB 53 passed, ENCODE conceded on third-party auditing and kill switches to secure the bill; however, they successfully secured inclusion of provisions covering internal deployments, a feature absent in earlier drafts.
- State-Level Regulatory Capacity: State laws in California and New York serve as "first steps" to build nimble, well-resourced regulatory capacity that can be activated quickly if federal action is delayed or insufficient.
- Illinois Legislative Precedent: Illinois passed legislation the week of the interview that includes third-party auditing requirements, demonstrating that state standards can rise and that successful state laws (like SB 53) provide a "role model" for other states to follow.
- Coalition Building Against Preemption: ENCODE helped form a diverse coalition including family-first conservatives, parents affected by AI addiction, and entertainment industry figures to successfully fight federal preemption of state AI laws.
- OpenAI Subpoena Incident: ENCODE's General Counsel, Nathan Calvin, received an individual subpoena from OpenAI in early 2025 regarding an amicus brief filed in the Musk v. Altman lawsuit concerning OpenAI's corporate restructuring.
- Strategic Response to Intimidation: ENCODE chose a diplomatic response to the subpoena to avoid burning bridges, aiming to maintain future cooperative relationships while highlighting the incident's implications for free speech and advocacy.
- OpenAI Policy Shifts: Following public pressure and internal dissent, OpenAI has shifted from opposing regulation to endorsing the Illinois safety bill (including audits) and calling for an international AI governance body.
- Political Funding Landscape: Anti-regulation super PAC "Leading the Future" raised $125 million for the midterms, while a counter-PAC backed by Anthropic is emerging; Ravenor notes that these "Big Tech" spending strategies may fail if they lack public salience and tailwinds.
- Midterm and 2028 Outlook: AI is becoming a top-tier issue for candidate differentiation in crowded primaries, with viral ads centering on AI data centers and job displacement, though it is not yet the primary motivator for voter turnout.
- Institutional Power Imbalance: Ravenor expresses concern that the Executive Branch may outpace the Legislative and Judicial branches in AI adoption, potentially concentrating power and upsetting the constitutional balance of power.
- Whistleblower Protections: ENCODE views whistleblower protections as a strategic "easy win" to codify voluntary industry practices, sequencing them to build political capital before pushing for more controversial measures like strict liability.
- Legislative Sequencing Strategy: The organization prioritizes codifying existing voluntary safety practices to realign industry incentives, reserving aggressive measures like financial liability for a future moment likely catalyzed by a specific crisis or "warning shot."
- Advice for Aspiring Advocates: Ravenor advises that successful AI advocacy requires toggling between "soldier" mode (persuasion and coalition management) and "scout" mode (rigorous, unsparing analytical thinking) to ensure actions are both politically effective and epistemically sound.
- Access to AI Safety Community: Ravenor highlights the unique approachability of the AI safety community as a resource for newcomers, encouraging them to reach out to experts and build networks despite their junior status.