Encode (formerly Encode Justice)

Theory of Change

Encode's theory of change operates on two levels:

Stated mechanism: Youth voices and bipartisan framing create political will for AI safety regulation at the state level, which then forces national standards through the "Brussels effect" (companies adopt the strictest jurisdiction's requirements globally). Sneha Revanur: "I don't want us to have to wait for an AI Chernobyl to start taking this seriously on a political level."

Operational mechanism: A small team of professional policy staff (Stanford JD, Senate Judiciary, CAIS Action Fund veterans) drafts and advocates for specific legislation, using the youth movement as a legitimacy base and the "moral high ground" of child safety issues to build bipartisan coalitions. Sunny Gandhi: "We have so much moral high ground, it's really hard to fight back against us."

The org has evolved from algorithmic justice (2020: fighting racial bias in bail algorithms) through AI ethics (2021-23: facial recognition, surveillance) to catastrophic risk (2024-present: AI 2030 platform includes x-risk language, CERN-like global safety institute proposals). The 2025 rebrand from "Encode Justice" to "Encode" with the tagline "pro-AI, pro-human, pro-progress, pro-innovation" reflects a strategic shift toward bipartisan positioning.

What They Do

Legislation -- the core output. Encode's record of concrete policy wins is unusual for its size:

• SB 1047 (CA, 2024): Co-drafted and co-sponsored alongside CAIS Action Fund. Would have required safety testing for frontier AI models. Passed both chambers; vetoed by Newsom.
• SB 53 (CA, 2025): Successor to SB 1047. Signed into law. Requires frontier AI transparency reports, whistleblower protections, catastrophic risk disclosures.
• TAKE IT DOWN Act (federal, 2025): Criminalizes nonconsensual AI deepfakes. Passed Senate unanimously. Bipartisan: co-sponsored by Ted Cruz and Amy Klobuchar. Encode was present at Trump's signing.
• NDAA FY2025 AI weapons provisions: First US law establishing guardrails on AI in nuclear weapons.
• RAISE Act (NY, 2025): Signed into law.
• Utah HB 286 (2026): AI transparency bill. Encode donated $10K to the Republican sponsor's campaign, commissioned polling, ran mass-texting.

Advocacy against corporate power. Filed amicus brief against OpenAI's for-profit conversion. Nathan Calvin was personally subpoenaed by OpenAI (sheriff's deputy at his home) demanding all his private communications about SB 53. OpenAI's own head of mission alignment (Joshua Achiam) publicly criticized the company's tactics in response.

Public education. AI Ethics workshops for 15,000+ students. AI 2030 platform (22 calls to action, signed by Bengio, Russell, Acemoglu, Kokotajlo). Plan for AI career readiness tool (launched Jan 2026).

Key People

Sneha Revanur -- Founder & President. 21 years old, Stanford senior. Founded Encode at 15 after discovering racial bias in criminal justice algorithms. TIME 100 AI youngest honoree. The public face and brand of the organization, but not a full-time employee ("still in college" per colleagues). Both parents are software engineers in Silicon Valley.

Nathan Calvin -- General Counsel & VP of State Affairs. Stanford JD/MPP. Previously senior policy counsel at CAIS Action Fund and Senate Judiciary Committee staffer. The most technically sophisticated policy mind on staff. Key drafter of SB 1047, lead on SB 53 and OpenAI opposition. His 80K Hours interview is the single most substantive public statement of Encode's policy thinking.

Sunny Gandhi -- VP of Political Affairs / Co-Executive Director. CS degree from Indiana. Led nuclear weapons AI provisions and defeat of federal preemption. Brings technical background combined with political operations.

Team has grown from 3 FTEs (October 2025) to 7 staff (March 2026). Membership claimed at 1,000-1,200 across 40+ states and 30+ countries, though chapter-level activity appears modest.

Money and Incentives

Legal structure: 501(c)(4) social welfare organization, fiscally sponsored by Future Incubator. The c4 status enables unlimited lobbying and political activity (including campaign donations) but means no public financial disclosures and no tax-deductible donations.

Known funding:

Total confirmed: ~$275K-$350K minimum. Actual budget likely $500K-$1M+ given 7 staff.

Independence claim: "We don't accept funding from corporations, foreign governments, or executives at top AI companies." This is unverifiable due to c4 status.

Key tension: Omidyar Network was the "primary source of funding" as of 2023 (per Politico). This concentration on a single progressive funder creates vulnerability to the accusation that Encode is a progressive front. However, funding from FLI and SFF (longtermist/x-risk community) and bipartisan policy achievements (TAKE IT DOWN with Cruz, Utah with Republican legislator) complicate this narrative.

Political spending: ~$10K campaign donation to Utah Rep. Fiefia, plus paid polling and mass-texting. This direct political activity is unusual for AI safety orgs.

Conflict of interest: Sneha Revanur's sister Swetha Revanur is Head of Engineering at Hebbia AI, backed by Peter Thiel, Google, and a16z -- entities that actively opposed SB 1047. Encode has not publicly disclosed this conflict.

What Others Say

Washington Examiner (March 2026): Investigated Encode as a "liberal organization influencing AI policy in a deeply red state." Documented progressive funding, staff backgrounds, campaign donations, and removed blog posts. Also investigated Secure AI Project (Nick Beckstead) collaboration. The strongest critical piece.

Zvi Mowshowitz (Oct 2025): Detailed analysis sympathetic to Encode, excoriating OpenAI's subpoena tactics. "Subpoenas are part of how both sides seek information... they don't assign fault" -- OpenAI CSO Jason Kwon's defense, which Zvi systematically dismantled. The AI safety community's reaction was overwhelmingly supportive of Encode: Helen Toner, Neel Nanda (DeepMind), Oliver Habryka, Michael Cohen, Dylan Hadfield-Menell all publicly criticized OpenAI.

Joshua Achiam (OpenAI head of mission alignment), publicly criticizing his own company: "At what is possibly a risk to my whole career I will say: this doesn't seem great... We can't be doing things that make us into a frightening power instead of a virtuous one."

Rachel Adjogah (Medium): Raised the Swetha Revanur / Hebbia conflict of interest, noting that Hebbia's backers (Thiel, a16z) actively opposed the regulation Sneha champions.

InfluenceWatch: Documented Encode's progressive roots, Omidyar funding, Cori Bush internship, and removed blog posts critical of ICE. Characterized Encode as pushing for "censorship" of AI.

Notably absent: no substantive criticism of Encode's theory of change or organizational model exists in any source. Nobody in the AI safety community or policy world has publicly argued that Encode's approach is wrong or ineffective.

What's Absent

• Total budget and detailed financials -- completely opaque due to c4 + fiscal sponsorship structure.
• Board of directors or governance structure -- none exists. All governance absorbed by fiscal sponsor.
• Conflict of interest disclosure -- the Swetha Revanur / Hebbia situation is unaddressed.
• Rebrand explanation -- no public accounting of why "Encode Justice" became "Encode" or why progressive content was scrubbed.
• Independent effectiveness evaluation -- no analysis of whether youth-led advocacy actually produces better AI policy outcomes.
• Member engagement reality -- "1,000 members" may mean signups rather than active participants.
• Post-graduation plans -- how does the org transition as its founder finishes college?

Recommended Reading

1. Nathan Calvin on 80,000 Hours -- The most substantive policy thinking from Encode. Calvin explains SB 1047 in depth, engages honestly with objections (open source, startups, federal vs. state), and reveals his views on catastrophic risk regulation. This is the single best source for understanding what Encode's professional staff actually thinks about AI policy. https://80000hours.org/podcast/episodes/nathan-calvin-sb-1047-california-ai-safety-bill/

2. Zvi Mowshowitz on OpenAI's lawfare against Encode -- Comprehensive analysis of the OpenAI subpoena incident including Nathan Calvin's full account, OpenAI CSO Jason Kwon's doubling down, and reactions from Helen Toner, Joshua Achiam, Gary Marcus, and others. Essential for understanding how the AI safety community views Encode. https://thezvi.substack.com/p/openai-15-more-on-openais-paranoid

3. Washington Examiner investigation -- The strongest critical perspective. Documents political donations, progressive funding ties, and the Utah advocacy operation. Essential for understanding how Encode is perceived by skeptics. https://www.washingtonexaminer.com/news/investigations/4493425/organizations-strong-ties-left-influenced-ai-deep-red-state/

4. Sneha Revanur in the Bulletin of the Atomic Scientists -- The most candid founder interview. Covers the founding story, evolution toward catastrophic risk, and relationship to Silicon Valley. https://thebulletin.org/premium/2024-01/interview-with-sneha-revanur-the-greta-thunberg-of-ai/