ControlAI: Legit UK Campaign Against Superintelligent AI

ControlAI is a campaigning group that wants development of superintelligent artificial intelligence (AI) prohibited. Its homepage argues that AI poses an extinction risk on par with nuclear war, and that prohibiting superintelligent AI can prevent it. One of its leaders, Connor Leahy, co-founded EleutherAI, which released AI language models including GPT-J-6B. Its UK registration is public on Companies House, but its funding isn't published.

What ControlAI Is and What It Wants

ControlAI campaigns to prohibit superintelligent AI, and a company registered in the United Kingdom (UK) under the name CONTROLAI has been active since 22 August 2023. Companies House lists it as a company limited by guarantee, with Andrea Miotti as a serving director. The group's About page gives its purpose as preventing 'the risk of human extinction posed by superintelligent AI'.

The About page says the group briefs lawmakers and engages the media, civil society and the public. It says ControlAI runs its work 'like a startup', with clear objectives and measurable goals. Its homepage frames the whole effort around one demand: prohibiting the development of superintelligent AI.

Searching the name has a problem. A search for 'control AI' mostly returns 'AI control', a research field about constraining and monitoring potentially misaligned systems. Redwood Research introduced that term. ControlAI's website, controlai.com, redirects to controlai.org.

The group wants development of superintelligent AI prohibited, and some of its leaders have worked on building AI models. The next sections cover who those leaders are and what their records show.

Who Runs ControlAI and How It Links to Conjecture

Andrea Miotti is listed as Founder and chief executive officer (CEO) on ControlAI's About page, read on 21 October 2026. Companies House also lists Miotti as a director of the UK company, appointed on 22 August 2023. Miotti's nationality is recorded as Italian.

Companies House lists three directors in total. Eva Christa Annelore Behrens, a German national, was appointed on 22 August 2023 and is still active. Nicolas Michel Pierre Moës, a Belgian national, was appointed the same day and resigned on 7 November 2023.

Connor Leahy is listed on the About page as United States (US) Executive Director. He isn't among the three UK directors on the register, so the UK filing doesn't name him as a director. Leahy is a German-American AI researcher and entrepreneur. He co-founded EleutherAI in July 2020 with Sid Black and Leo Gao, initially to replicate GPT-3.

While co-leading EleutherAI, the group released The Pile in December 2020, GPT-Neo in March 2021, GPT-J-6B in June 2021 and GPT-NeoX-20B in February 2022. Leahy co-founded Conjecture, an AI safety company, in March 2022 and was its CEO until 2026.

In March 2026, Leahy announced that Conjecture was winding down and that he was joining ControlAI as US Director, according to Wikipedia. He had previously advised ControlAI. The About page uses the title US Executive Director, so the two sources give different titles.

The Conjecture link runs wider than Leahy. The About page lists Gabriel Alfour as an adviser. It describes him as an entrepreneur who co-founded Conjecture and helped Andrea Miotti found ControlAI.

The About page names 16 team members. They include Leticia García Martínez, UK Parliamentary Engagement Lead, Sophie Toura, Head of UK Operations, and Grace Gonzales, Media Engagement Lead. A list of names isn't a headcount, and no staff headcount is published.

Leahy's public record predates ControlAI. He signed an open letter from March 2023 calling for a pause on training AI systems more powerful than GPT-4, and the Statement on AI Risk from May 2023. At the inaugural AI Safety Summit in November 2023, he warned that if uncontrolled AI surpasses humans, 'the future belongs to them, not us.' Wikipedia also records that he worked with ControlAI to advocate for a pause on artificial general intelligence (AGI) development.

Our Connor Leahy profile covers his wider record, and our AGI explainer covers the term.

What ControlAI Campaigns For and Who Has Signed On

ControlAI runs a statement and supporter campaign on extinction risk from AI in the US and UK. It also runs a separate call for action on superintelligence in Canada, according to its About page. The central demand across these campaigns is a ban on building superintelligent AI.

On legislation, ControlAI says it consulted on what it describes as the first US Congress bill to ban superintelligent AI. It supports a UK bill to ban superintelligent AI, and its About page says over 85 UK lawmakers back that bill.

The group's reach figures are its own. Its About page says ControlAI has briefed over 200 US congressional offices in just over a year. It also says it has briefed over 400 lawmakers across the US, UK, Canada and Germany over two years. The homepage gives the same 400-plus briefing figure.

The lawmaker-support numbers conflict, because they measure different things. The homepage says 170+ UK lawmakers support the campaign and 35+ Canadian lawmakers do. A homepage banner said 85+ cross-party Members of Parliament (MPs) and Peers had called on the Prime Minister to evaluate and adopt the bill, after an earlier figure of 70+. Read each figure with its page and date, and don't add them together.

On supporters, the About page claims about 150,000 supporters have sent over 200,000 messages to lawmakers, mostly in the US. The homepage says more than 148,000 people read its weekly newsletter. All of these counts are self-reported by ControlAI and were read on 21 October 2026, so they have no independent check.

Named UK lawmakers on the homepage include The Viscount Camrose, former Minister for AI, The Lord Clement-Jones CBE, The Rt Hon Anneliese Dodds MP and The Rt Hon Sir John Whittingdale OBE MP. Non-legislators named as endorsers include Mary Robinson, Lord Browne of Ladyton, Anthony Aguirre, Roman Yampolskiy, Yuval Noah Harari, Stuart Russell, Dan Hendrycks and Steven Adler.

No bill ControlAI backs has become law. ControlAI says it publishes a 2025 Impact Report and a list of past campaigns on its own site. The About page is the starting point for that self-reported record.

Readers comparing ControlAI with other groups fighting AI can start with our guide to who is fighting AI and our PauseAI and StopAI guide.

How ControlAI Is Funded and Whether It Takes Public Donations

ControlAI doesn't publish a donor list. Neither its About page nor its homepage names a single funder. Its homepage also says it accepts only substantial donations, with exceptions in special cases.

The homepage has a 'For Donors and Partners' section. It invites donation and funding enquiries through a contact form. Because it accepts only substantial donations, the site isn't running a public small-donor appeal, and it doesn't describe a small-donor route.

Its German lobby-register entry for CONTROL AI US INC has no funding figures. The entry reads 'Keine Angaben, da noch kein Geschäftsjahr vollständig abgeschlossen wurde', which translates as 'No information, since no full fiscal year has been completed yet.' The entry doesn't give a figure for lobbying staff in full-time equivalent (FTE) terms, or for annual lobbying spending.

Reporting from the Washington Post, as summarised by AI Weekly, says Jaan Tallinn, a Skype founding engineer, is the only publicly named funder. That attribution rests on the AI Weekly summary rather than on the Washington Post article itself. No donation amount is published for Tallinn or anyone else.

The UK entity is a company limited by guarantee, not a registered charity. No Charity Commission accounts were available to check. Companies House lists its registered business activity as public relations and communications activities, under Standard Industrial Classification (SIC) code 70210.

No revenue, expenditure or asset figure of any kind could be verified for ControlAI, in the UK or the US. The blank figures in the lobby register are a public gap.

Is ControlAI Legit?

Yes, as a registered company. The UK entity is real and active, with named directors and a registered office. Its money, its US tax status and the link between its two entities are not verified here.

The UK record checks out on these points, all from Companies House:

• Company name: CONTROLAI, company number 15088415, status Active.
• Incorporated 22 August 2023, with a registered office at 201-211 Borough High Street, London, England, SE1 1JA.
• Type: private limited company by guarantee without share capital, using the 'Limited' exemption.
• Business activity: public relations and communications activities (SIC 70210).
• Previous name: SECURE FUTURE RESEARCH LTD, in use from 22 August 2023 to 10 September 2026.

CONTROL AI US INC is a separate US entity. It is registered in the German Bundestag lobby register under number R007795. The entry declares its legal form as a 501(c)(4) organization, which the entry calls a Social Welfare Organization, at an address in Washington, DC. It was first filed on 19 January 2026 and last updated on 16 March 2026.

The entry names Andrea Miotti, CEO, as both authorised representative and the person directly conducting interest representation. Its declared fields of interest include data protection, communication and information technology, science and research, and artificial intelligence.

The gaps are on the public record too:

• The relationship between the UK company and CONTROL AI US INC isn't stated in any record checked.
• The UK company has no charity registration, so no charity accounts are available.
• No Internal Revenue Service (IRS) record for CONTROL AI US INC was found, so its 501(c)(4) status rests on its own German filing.
• The previous name runs until 10 September 2026, though the group had been campaigning as ControlAI well before then. The filings don't explain the gap.
• The name 'Direct Institutional Plan' doesn't appear on ControlAI's current site.
• No staff headcount is published.

To check the UK registration yourself, open the officers page for company 15088415. It lists each director with appointment and resignation dates.

Criticism of ControlAI

The main structural criticism is that ControlAI's leaders and advisers come from Conjecture. Leahy co-founded it, and Gabriel Alfour co-founded it too and helped Andrea Miotti found ControlAI. Critics read that as a small, self-referential network campaigning for a global prohibition.

A 2023 post on the Effective Altruism (EA) Forum and LessWrong, linked via GreaterWrong, argued that Conjecture had 'misrepresented themselves multiple times'. It recommended against further funding until the concerns were addressed. Other commenters disputed parts of it as outdated. That critique is about Conjecture, not ControlAI.

The funding criticism is the conflict in the named funder. Reporting says Jaan Tallinn, the only publicly named funder, was an early backer of one of the frontier AI labs. ControlAI argues companies of that kind should not be permitted to build superintelligence.

The transparency criticism follows from the records. ControlAI publishes no donor list, its German lobby entry has no spending figure, and the UK company has no charity accounts. The scale and sources of its lobbying money can't be checked from public filings.

A critic has questioned whether ControlAI supplied drafting language for US legislation. The same critic asked whether the funder's ties to a frontier lab were adequately disclosed. The questions come from an advocacy social-media thread, not from reporting, and they remain unresolved.

The case for ControlAI rests partly on its endorsers. Stuart Russell, Yuval Noah Harari, former Irish President Mary Robinson and former UK Defence Secretary Des Browne are named endorsers with no stake in the group, according to its homepage and About page. Its lawmaker support is cross-party.

The argument about extinction risk itself is covered in our AI doomers explainer. The group doesn't name its donors, and the filings leave out spending figures.

Sources

• ControlAI, About page
• ControlAI, homepage
• Companies House, company 15088415 (CONTROLAI)
• Companies House, officers of company 15088415
• German Bundestag lobby register, entry R007795 (CONTROL AI US INC)
• Wikipedia, Connor Leahy
• AI Weekly, summary of Washington Post reporting on ControlAI
• GreaterWrong, critiques of prominent AI safety labs: Conjecture
• Redwood Research, AI control