iiinova — where your digital self keeps evolving

Gillian Hadfield @ghadfield.bsky.social · 09/10/2026

I've been at this for almost a decade, first as regulatory markets and now with Fathom as the IVO model. We all thought we had more time. We didn't, but luckily we have something concrete and shovel ready for the moment. What we need now is for governments to act on it while the window is open.

000

Gillian Hadfield @ghadfield.bsky.social · 07/10/2026

People ask me what happened. We've been building AI for 80 years. The key thing: humans stopped writing the rules. With machine learning we show the machine the data and it writes its own program. That moved us back a step from controlling these systems, and now AI is helping build the next AI.

041

Gillian Hadfield @ghadfield.bsky.social · 03/10/2026

I've been thinking about how to govern powerful AI for about ten years. I wasn't all that worried until the last six months. More worried in the last two months. More worried in the last week. Regulatory markets and IVOs are what we should be doing. I'm hoping it's not too late.

141

Gillian Hadfield @ghadfield.bsky.social · 30/09/2026

We have a handful of expert independent groups capable of testing whether frontier models actually have the controls the companies say they do. Little green shoots. We need mighty oaks, and fast, and that means money and brains going into an independent verification sector.

042

Gillian Hadfield @ghadfield.bsky.social · 29/09/2026

The first AGI Governance Fellowship cohort on their last day at Johns Hopkins, after their group project presentations. Three weeks of hard questions on the institutions we'll need for powerful AI. A great group. Thanks to the fellows, @sethlazar.org, @nickacaputo.bsky.social and all who joined us.

051

Gillian Hadfield @ghadfield.bsky.social · 28/09/2026

Letting evaluators into AI labs isn't oversight when the lab picks them, sets their access and can show them the door. To do the job right, evaluators need serious oversight. Who decides they're qualified? What keeps them independent? What happens if they do the job badly?

141

Gillian Hadfield @ghadfield.bsky.social · 26/09/2026

I spoke to Salma Abdelaziz on CNN International about how to slow down AI when the US and China are racing. The people closest to the technology are the ones asking for it. Slowing down doesn't mean stopping. It means making sure a system is safe enough before it goes out.

020

Gillian Hadfield @ghadfield.bsky.social · 25/09/2026

The recording of my Next Conversations panel with Stuart Russell and @deanwb.bsky.social at the Hopkins Bloomberg Center is up. We don't agree on everything, but we agree the gap between what these systems can do and the tools we have to keep them in check is widening fast.

101

Gillian Hadfield @ghadfield.bsky.social · 23/09/2026

His EO tasks us to explore roles for IVOs, which builds on my work with Fathom, in contexts that include assessing the efficacy of a kill switch and the adequacy of frontier labs' safety frameworks. Recs due Nov 16. buff.ly/qOd9tBd

buff.ly

Governor Newsom announces world-leading experts to deliver on his AI executive order, including advancing creation of a “kill switch” | Governor of California

Official website of the State of California

010

Gillian Hadfield @ghadfield.bsky.social · 23/09/2026

Thank you to Governor Newsom for asking me to join the group of experts advising California on AI safety and security governance.

150

Gillian Hadfield @ghadfield.bsky.social · 18/09/2026

I spoke to @amitkatwala.bsky.social at MIT Tech Review about DeepMind's new swarm experiment, where agents cheated and others blew the whistle. Official channels to talk may have contributed to enforcement. Alignment is institutional, not (just) dispositional. buff.ly/K2e5p0a

buff.ly

AI agents blew the whistle on their cheating colleagues

Swarms of AI agents could supercharge scientific progress or wreak havoc. New research from Google DeepMind suggests that peer pressure could keep them in line.

251

Gillian Hadfield @ghadfield.bsky.social · 11/09/2026

I spoke to TIME about the Hugging Face incident and AI agents. If you said we're building new members of a group, our group, you'd build them differently than you're building them now. Alignment is not just an engineering problem. It's fundamentally institutional. buff.ly/YO18tlS

162

Gillian Hadfield @ghadfield.bsky.social · 11/09/2026

California will now designate independent verification organizations, outside experts qualified to assess the risks of AI models. Newsom signed SB 813 yesterday, the biggest step yet toward the independent verification sector we need. Thank you to @senmcnerney.bsky.social and Fathom. buff.ly/RHxB25H

gov.ca.gov

Governor Newsom signs first-in-the-nation AI safeguards to protect Californians, calls on the federal government to do its part | Governor of California

Official website of the State of California

020

Gillian Hadfield @ghadfield.bsky.social · 10/09/2026

Delighted to welcome Brigit Goebelbecker as the new CEO of @coop-ai.bsky.social. Lucky to have her at such a critical time for cooperative AI. Looking forward to working with Brigit, Research Director Lewis Hammond and everyone at the Foundation. Welcome! buff.ly/z6OCcGd

cooperativeai.com

CAIF Appoints First CEO

We are pleased to announce the appointment of Brigit Goebelbecker as Chief Executive Officer, starting October 2026.

030

Gillian Hadfield @ghadfield.bsky.social · 05/09/2026

Fantastic opportunity for ambitious ML researchers, and a really great group of people to work with. Up to 12 Fellows across Oxford, UCL and Imperial, with full academic freedom. Deadline is Sept 15. Please spread the word in your network! buff.ly/nfOmBli

my.corehr.com

Job Details

010

Gillian Hadfield @ghadfield.bsky.social · 04/09/2026

The AI investigating the Hugging Face hack took the rogue agents' side. METR used similar models to read the transcripts, and its chief scientist called them "very credulous." Same finding in Talk Isn't Always Cheap: agents swap reasoning and flip from right to wrong. buff.ly/XtuO2GO

bsky.app

Dylan Freedman (@dylanfreedman.nytimes.com)

OpenAI voluntarily let three researchers from A.I. safety nonprofits investigate how its rogue A.I. agents hacked Hugging Face, leading to the most comprehensive account yet of the alarming incident…

130

Gillian Hadfield @ghadfield.bsky.social · 03/09/2026

@sethlazar.org, @nickacaputo.bsky.social, and I are building Governing the AI Transition, a new Johns Hopkins SGP initiative to help society navigate the transition to powerful AI. We're hiring a Director to build it with us. If you're ready to roll up your sleeves, please apply! buff.ly/zPa3bFu

hiring.jhu.edu

GAIT Director (School of Government and Policy) | Johns Hopkins University

Support faculty strategic leadership of GAIT by executing large-scale projects in alignment with the School's strategic goals. Ensure initiatives meet sponsor deliverables, compliance requirements,…

061

Gillian Hadfield @ghadfield.bsky.social · 03/09/2026

Yesterday I was on Capitol Hill with Fathom briefing House staff on the FRONTIER Act, the bipartisan bill from Rep. Obernolte and Rep. Trahan that would license independent experts to verify the safety of frontier AI. Full room, and a clear sense that Congress needs to move on this. buff.ly/KqNFbDG

linkedin.com

https://www.linkedin.com/feed/update/urn:li:activity:7500939501192585217

141

Gillian Hadfield @ghadfield.bsky.social · 01/09/2026

Over half of internet traffic is now non-human. With Dan Hendrycks and Leo Wu, I look at agent IDs, deployment cards, personhood, and payments. It mostly comes down to how much we let agents do and how much oversight we keep. buff.ly/AQjYo3k

ai-frontiers.org

We Need Better Infrastructure to Govern AI Agents

Gillian Hadfield, Aug 27, 2026 — Society is not prepared for a flood of agents. We need new protocols and standards, such as Agent ID, to make agents accountable to our legal and financial systems.

085

Gillian Hadfield @ghadfield.bsky.social · 31/08/2026

Roughly 700 OpenAI agents hacked Hugging Face. METR and Redwood's independent investigation took three researchers and six days. IVOs exist to make that scrutiny routine. Last night California became the first state to start building the IVO sector. #AIGovernance #IVO buff.ly/3aJwdz8

fathom.org

California Legislature Overwhelmingly Passes Fathom-Sponsored Bill to Spur Independent Verification of AI Safety - Fathom

Building solutions to navigate the transition to a world with AI

020

Gillian Hadfield @ghadfield.bsky.social · 26/08/2026

Thanks to The National Law Review and Wickard for naming me to their 2026 Top 50 Legal Innovators in Academia. Congratulations as well to the other honorees, a strong group working across AI, law, and legal education. natlawreview.com/article/2026...

natlawreview.com

The 2026 Top 50 Legal Innovators in Academia

The National Law Review and Wickard are proud to announce the Top 50 Legal Innovators in Academia for 2026, a national recognition honoring the educators, administrators, researchers, and academic…

030

Gillian Hadfield @ghadfield.bsky.social · 25/08/2026

A crib gets a safety sticker because someone independent tested it first, so parents don’t have to. We actually have that for almost everything else in kids’ lives. Ohio’s HB 628 would license independent verifiers to do it for AI. #AIGovernance #IVO www.clermontsun.com/2026/08/19/l...

clermontsun.com

Letter to the Editor: AI at home

<p>When my kids were babies, I checked the crib for the safety sticker and the car seat for recalls. Someone I trusted had already tested those things and decided they were safe.</p>

040

Gillian Hadfield @ghadfield.bsky.social · 12/08/2026

Thanks to Knowledge Networks and Regulating AI for including me in this year's AI Policy 100. Congratulations as well to the others on the list working in this critical domain. natlawreview.com/press-releas...

natlawreview.com

Knowledge Networks & Regulating AI Introduces The AI Policy 100, Honoring the Most Influential Voices in AI Governance

34 New Articles

000

Gillian Hadfield @ghadfield.bsky.social · 10/08/2026

Andrew Freedman [tag] and I discuss in Fortune how the Obenornolte-Trahan bipartisan proposal in Congress, the FRONTIER Act, would begin building an effective ecosystem of independent verifiers that would help ensure events like this can't be kept secret in the AI industry. buff.ly/na4Krvo

fortune.com

AI labs shouldn't be allowed to grade their own homework | Fortune

We know about AI labs' hacking failures only because the companies involved chose to tell us.

030

Gillian Hadfield @ghadfield.bsky.social · 10/08/2026

We only learned about OpenAI and Anthropic agents hacking into secure systems because the companies chose to tell us. But if Boeing discovered a dangerous problem with one of its aircraft, it wouldn't get to keep that information to itself. Drug companies are obligated to report adverse events.

111

Gillian Hadfield @ghadfield.bsky.social · 31/07/2026

2/ I and others have been working on the problem of how to build such infrastructure for ten years, including participating in dialogues on AI safety with Chinese academic colleagues during the past three. Here are my suggestions: buff.ly/VNLRbFx

010

Gillian Hadfield @ghadfield.bsky.social · 31/07/2026

1/ The Pacing the Frontier letter calls on the US government to support an international effort to build the technical and governance tools needed to protect our option to pace AI development. bsky.app/profile/yosh...

bsky.app

Yoshua Bengio (@yoshuabengio.bsky.social)

1000+ scientists at frontier AI companies are speaking out to warn that the current commercial race leads to unacceptable security risks. I agree with their call for an international effort to…

120

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

6/ Insurers regulate. They always have. Through pricing, they decide who gets to participate and on what terms, and they do it by pricing rather than by rule. Give that verification a public outcome to measure against, and coverage comes back. buff.ly/QjEcfh8

buff.ly

Underwriting the Agent Economy: The Blueprint for an AI Insurance Stack

Insurance has been the quiet enabler of major economic and technological developments. This report describes the eight-component insurance infrastructure stack needed to keep AI agents insurable.

020

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

5/ The report points to a middle ground. Government issues revocable licenses to organizations that audit AI against standards the market develops. That is close to the model Jack Clark and I set out in Regulatory Markets, which the report cites.

100

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

4/ Underwriters Laboratories was founded in 1894 because electricity was setting buildings on fire. Insurers wrote the standards, and the mark became the price of being insured. But AI agents move faster than a standard can be written into law.

110

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

3/ So the market is leaving. Insurers are issuing exclusions that strip generative AI out of general liability policies. As the report puts it, that is a "market exit, with no clear path to re-entry."

100

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

2/ A new report from the Artificial Intelligence Underwriting Company found that liability from frontier AI agents is largely unpriced and invisible. Insurers have 90% of their exposure to AI-agent risk sitting in silent coverage.

110

Gillian Hadfield @ghadfield.bsky.social · 27/07/2026

1/ Air Canada had to honor a discount its chatbot invented. The liability caused by AI agents is landing on policies written by an insurance industry that never planned for them.

140

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

8/ This is the first in a short series from a new PNAS piece, where I set out fixes for the same underlying gap. doi.org/10.1073/pnas...

000

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

7/ And whether it's the developer who built the agent or the person who sent it off, a way for them to be held to account is required to avoid serious disruption to our markets.

100

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

6/ It just means a public and traceable identity, legally tied to a person or entity that can be held accountable for what an agent does on their behalf.

100

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

5/ That record is what makes them accountable, and it's something we're lacking when it comes to AI agents. This isn't about special rules for AI or restricting how AI agents interact with the market; it's having them follow the same guidelines as every other participant.

210

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

4/ We call these actors corporations, and when a company signs a contract it cannot bring to term, no one has to hunt down their shareholders. Companies have registered agents and addresses on file where papers can be served.

100

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

3/ We focus on what the rules say and miss the previous step. How do we even enforce the rules we make? Society already has a solution for how we register artificial actors that act in markets on someone's behalf.

100

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

2/ How do you know who to sue, or report? Right now there's no reliable way to answer who an AI agent reports back to, because we haven't built the legal infrastructure the courts would use to identify the actor behind an agent.

100

Gillian Hadfield @ghadfield.bsky.social · 23/07/2026

1/ AI agents that can sign contracts on your behalf, hire employees, set prices, and move your money around are being heavily invested in by AI companies. But what I want to call attention to, what happens if an agent sells you faulty goods or runs off with your deposit?

150

Gillian Hadfield @ghadfield.bsky.social · 13/07/2026

I joined over 200 economists and AI researchers in signing "We Must Act Now" a statement on AI's transformation of the economy. AI could reshape the economy at unprecedented speed. The opportunities are enormous and so are the challenges. We need to start preparing our institutions.

wemustactnow.ai

We Must Act Now: A Statement on AI’s Transformation of the Economy

1100

Gillian Hadfield @ghadfield.bsky.social · 10/07/2026

4/ Right now the decision on what counts as safe sits with private companies, not with anyone democratically accountable. Jack Clark and I called this the democratic deficit in our regulatory markets paper. buff.ly/5S8JRaD

gillianhadfield.org

REGULATORY MARKETS:THE FUTURE OF AI GOVERNANCE

120

Gillian Hadfield @ghadfield.bsky.social · 10/07/2026

3/ Under SB315, a company could meet every one of its standards, get a full sign-off from the auditor, and still release a model that proves to be dangerous.

100