Centre for Long-Term Resilience (CLTR)

Theory of Change

CLTR is a UK think tank founded in November 2019 by two former senior civil servants (Angus Mercer and Sophie Dannreuther) to translate academic extreme-risk research into actionable UK government policy. Their stated mechanism: provide governments with "concrete policy recommendations" and "the extra hands it needs" on AI, biosecurity, and risk management, operating as insiders who understand Whitehall's machinery from the inside.

Angus Mercer's founding rationale (2023): "Extreme risks -- and government risk management generally -- felt to me like an extremely important policy area. It also seemed to be highly neglected -- there was every incentive for governments to not give this area the attention it deserved, and instead to focus on issues that were more pressing in the short term."

CLTR explicitly rejects strong longtermism. Mercer: "When we talk about governments taking a longer-term view... we're very much not talking about taking a thousand or million-year view. We're talking about the concrete, actionable steps that governments can take to ensure that over the next five, 10, 20 years, we become much more resilient." The working time horizon is 5-20 years, not the existential risk timescale.

A secondary theory of change is the "London effect": the UK can set global regulatory blueprints that other nations adopt, leveraging the UK's convening power, close US relationship, and AISI's technical capabilities.

What They Do

Three policy pillars: AI Policy, Biosecurity Policy, Risk Management Policy. Each has a director-level lead and dedicated staff. The most distinctive recent outputs:

Loss of Control Observatory (2025-2026): The first systematic real-world monitoring of AI scheming behaviors. Analyzed 180K+ transcripts from X, identified 698 scheming-related incidents over 5 months, found a 4.9x increase. Novel OSINT methodology funded by the UK AISI Challenge Fund. No other organization does this.

AI governance policy work: Detailed recommendations for governing loss of control risk (Feb 2026) -- four proposals including adding AI loss of control to the National Risk Register, designating DSIT as Lead Government Department, publishing an AI Security Strategy, and introducing emergency powers. A "preparedness framework" for the UK AI Bill (Sep 2025) modeled on the UK Biological Security Strategy. Analysis of advancing UK global leadership in frontier AI governance (Dec 2025).

Risk management translation: Report proposing Three Lines Model (private sector risk governance best practice) for frontier AI companies. This cross-pollination from corporate risk management to AI safety is genuinely distinctive.

Claimed policy wins (self-reported, 2023): Input adopted in the MoD AI Strategy; expertise used in the UK Biosecurity Strategy refresh; NSRA time horizon expanded from 2 to 5 years; involvement in Bletchley Park AI Safety Summit preparations. None independently verified.

30+ publications to date, with output accelerating in 2025-2026. Five parliamentary evidence submissions. Commissioned public polling on extreme risk attitudes from Public First. Convenes workshops with DSIT and participates in frontier AI regulation roundtables with the DSIT Secretary of State.

Key People

Angus Mercer, Founder & CEO. Lawyer, former Head of External Affairs at DFID, MSc Global Governance & Diplomacy (Oxford), CSER Research Affiliate. Director since incorporation in 2019. No public podcast interviews or long-form candid discussions found -- for a think tank CEO, unusually absent from public discourse.

Jess Whittlestone, Senior Adviser (formerly Head of AI Policy). PhD Behavioural Science (Warwick), first-class Maths & Philosophy (Oxford). Former senior research fellow at CSER and LCFI (Cambridge). TIME 100 AI 2023. CLTR's most publicly visible figure. Explicitly skeptical of AI company claims: "AI companies keep saying the big societal benefits are just around the corner, if you keep letting us push forward. I am somewhat skeptical of that." Bridges near-term and long-term risk communities.

Team: ~20 staff, up from a 2-person founding team in 2019. Policy-focused -- no ML researchers or technical AI safety researchers. Recent hiring for Director of AI Policy at GBP 100K+ salary. Key hires from FHI, GovAI, Australian government, and corporate risk management (Cathay Pacific CRO).

Notable departure: Co-founder Sophie Dannreuther resigned from the board December 31, 2024. No public announcement or explanation. She is no longer on the team page.

Money and Incentives

Total known funding: >$9.5M+ (excluding SFF amounts from 2024-2025 and Open Phil matching). Detailed budget breakdowns are not publicly available.

Funding concentration: >80% of known funding comes from three EA-adjacent sources (Open Phil, SFF/Tallinn, FLI). Despite explicit organizational distancing from EA since April 2023, CLTR's financial base is overwhelmingly EA-adjacent.

Business model: Grants and donations. No product revenue, no government contracts (though AISI Challenge Fund blurs this line). Claims to reject government funding as a matter of policy (to maintain independence). Director-level salaries at GBP 100K+ with generous benefits (GBP 5K wellbeing budget, GBP 3K L&D, private health insurance, 7% pension match).

Budget trajectory: ~GBP 903K (2022) growing to an estimated GBP 2.5-3.5M/year by 2025-2026 (based on the Open Phil base of ~GBP 1.3M/year plus matching, SFF, and other sources). Novara Media claimed >GBP 2M/year in 2023.

Lab ties: None identified. No AI company funding, no compute dependencies, no lab board seats. This is a genuine strength for policy credibility. Jess Whittlestone explicitly positions herself as free from industry conflicts of interest.

Financial opacity: CLTR operates as Alpenglow Group Limited, a private company limited by guarantee -- not a registered charity. This means no Charity Commission oversight, no public accounts, no mandatory financial disclosure. This is unusual for a UK think tank of this size. Most comparable UK think tanks (IPPR, Chatham House) are registered charities with public finances. A second company ("The Centre for Long-Term Resilience Ltd") was created in April 2021 and dissolved in August 2022 with no public explanation.

What Others Say

Novara Media (April 2023) -- the only substantive external criticism found. Matteo Tiratelli describes CLTR as part of EA's "revolving door" in UK politics: "a thinktank set up by two former civil servants which has received over GBP 2M a year in funding, a substantial proportion of it from Tallinn and Sam Harris." CLTR is named alongside CSER, FHI, and APPG for Future Generations as channels of EA political influence. The article critiques EA as drawn to "neoliberal policy solutions." The Sam Harris funding claim is unverified elsewhere.

Founders Pledge (September 2023) -- recommends CLTR as effective. "CLTR has a proven track record of influencing UK policy on extreme risks." Notes UK government policy is "critical for reducing global extreme risk" given London's role as an AI hub.

TIME (September 2023) -- Jess Whittlestone named TIME 100 AI. Profile reveals her nuanced position: takes extreme risks seriously but is skeptical of industry, concerned that "the more extreme risks" are getting disproportionate attention relative to "more diffuse societal harms -- like inequality, or disinformation."

EA community: Zero presence. No EA Forum posts, no LessWrong discussion. CLTR operates entirely outside EA community spaces.

What's Absent

• Public financial statements. Unusual for a think tank of this size. The company limited by guarantee structure avoids the transparency requirements of UK charity registration.
• Independent effectiveness evaluation. No external audit of whether CLTR's policy recommendations are technically sound, or whether claimed policy wins are genuinely attributable to CLTR's work.
• Explanation for co-founder's departure. Sophie Dannreuther left in December 2024 with no announcement. Second unexplained board resignation (Gabriella Overodder, September 2025) within 9 months.
• Public candid interview with CEO. Angus Mercer has no podcast appearances, no long-form interviews despite leading a think tank whose theory of change depends on persuasion.
• Technical AI safety expertise. No ML researchers or alignment researchers on staff. CLTR depends entirely on external experts for the technical substance of its AI policy recommendations.
• Conflict of interest policy. No published policy, despite criticizing industry conflicts of interest and having advisory council members from the government departments CLTR seeks to influence.

Recommended Reading

1. "Why we started CLTR" (April 2023) -- https://www.longtermresilience.org/why-we-started-cltr/ -- The most candid source. Both founders explain motivations, EA inspiration, and explicit reservations about EA/longtermism. Read both Angus and Sophie's sections.

2. "Effective Altruism Is Infiltrating UK Politics" (Novara Media, April 2023) -- https://novaramedia.com/2023/04/07/effective-altruism-is-infiltrating-uk-politics-and-its-time-we-woke-up/ -- The strongest external criticism. Names CLTR as part of EA's "revolving door" in UK politics. Puts CLTR in the context of broader EA political influence.

3. TIME 100 AI: Jess Whittlestone (September 2023) -- https://time.com/collection/time100-ai/6309024/jess-whittlestone/ -- Reveals the intellectual stance of CLTR's most prominent figure: skeptical of AI companies, free of industry conflicts, bridging near-term and long-term risk communities.

4. "How the UK Government can govern the risk of loss of control" (February 2026) -- https://www.longtermresilience.org/reports/how-the-uk-government-can-govern-the-risk-of-loss-of-control/ -- CLTR's most substantive recent AI policy work. Four concrete recommendations showing deep knowledge of UK governance.

5. "Scheming in the Wild" (March 2026) -- https://www.longtermresilience.org/reports/scheming-in-the-wild/ -- Loss of Control Observatory findings. Novel OSINT methodology. 698 incidents. The most distinctive thing CLTR does.