ResearcherSource
A. Agrawal / S. Ebadian / L. Hammond
Authors, 'The Multi-Agent Off-Switch Game' (AAMAS 2026)
Extends the off-switch game to multi-agent settings.
Recent activity
[2026-05] The Multi-Agent Off-Switch Game (AAMAS 2026)132 profiles
ResearcherSource
Aaron Kingsley Clark
Researcher, School of Mathematics & Computational (per arXiv)
'The Veto Variable' (arXiv:2609.00109): human override as a goal-independent cost term.
Recent activity
[2026-09] The Veto Variable (arXiv:2609.00109)Promoter - expert-journalist matchmaking and safety communicationsSource
AI Safety Communications Centre (AISCC)
Organization - connects journalists with AI safety experts
Listed on the AI Security & Safety portal as an AI-policy nonprofit; related coverage of an AI-safety guidance tool for journalists (CUNY, May 2026).
Recent activity
New AI Safety Tool expands critical safety guidance for journalists (related coverage)
Researcher
Alessandro Facchini
Associate Professor (SUPSI), Swiss AI Lab IDSIA (USI-SUPSI)
Co-author of the 2025 off-switch signalling-game paper.
Source
Researcher
Alessio Benavoli
Professor, Universita della Svizzera italiana (USI)
Lead author of 'The AI off-switch problem as a signalling game' (PMLR 290, SIPTA 2025).
SourceAnalystSource
Alex Miller
Reporter (politics)
Fox News politics reporter covering the Senate dispute over a proposed AI kill-switch requirement.
Recent activity
Rand Paul clashes with fellow Republican over AI kill switch
Promoter
Alex Sobel
Member of Parliament, Leeds Headingley (Labour)
Plans to introduce the UK Artificial Intelligence Security Bill (8 Sep 2026, backed by ControlAI) to prohibit development of uncontrollable superintelligent AI - would make the UK the first G7 country to do so; earlier Commons amendment attempt in May 2026 failed.
SourceAnalystSource
Alexander Bolton
Senior staff writer (Congress)
The Hill congressional reporter covering the Senate-floor clash over emergency shutdown requirements for advanced AI.
Recent activity
Republicans clash on Senate floor over AI kill switchAnalyst
Alexandra Agraz
Writer
Legal writer covering California's executive order and emerging AI shutdown and verification requirements.
Source
AnalystSource
Alexis Dufresne
Writer / Analyst
Writer analyzing proposals for independently verifiable AI kill-switch requirements.
Recent activity
Anthropic co-founder backs verifiable AI kill-switch rules
BuilderSource
Anand Kannappan
Co-founder & CEO
Co-founded Patronus AI, which provides evaluations, guardrails, and scalable supervision for agentic systems.
Recent activity
Patronus unveils scalable supervision for agentic systems
Researcher
Andrew Garber
Researcher, UC Berkeley CHAI
Co-author, The Partially Observable Off-Switch Game (AAAI-25).
Source
BuilderSource
Ankur Shah
Co-founder & CEO
Leads Straiker, which develops agent discovery, adversarial testing, and runtime protection for enterprise AI agents.
Recent activity
[2026-10] Straiker wins 2026 CyberSecurity Breakthrough Award (AI security solution of the year)
Promoter
Anna Paulina Luna
U.S. Representative (R-FL)
Cosponsored H.R. 9917; calls a kill switch necessary to 'protect humanity'; developing separate legislation with Ro Khanna for a human interface and kill switch.
Source
CriticSource
Aura Salla
MEP (EPP, Finland); ITRE
'Europe cannot keep building its tech stack on access that can be switched off overnight by a foreign government.'
Recent activity
[2026-02] Watchdog organisations call to withdraw Salla's Digital Omnibus rapporteur role (Corporate Europe)
Critic
Bart Groothuis
MEP (Renew Europe, Netherlands); ITRE; ex-Chips Act negotiator
'Europe needs its own LLMs and open weight models or face digital colonisation.'
Source
Builder
Ben Kliger
Co-founder & CEO
Leads Zenity, an enterprise security platform focused on agent discovery, runtime boundaries, and protection from hijacked or rogue actions.
Source
BuilderSource
Beth Barnes
Founder & CEO
Founded METR, which evaluates autonomy, self-replication, sabotage, and other capabilities relevant to release gates and loss-of-control risk.
Recent activity
METR and the AI safety talent shortage
Promoter
Brad Carson
President, Americans for Responsible Innovation
'Advanced AI models should never be deployed without a reliable off switch.'
SourcePromoter
Brendan Steinhauser
CEO, The Alliance for Secure AI
'Humans remain the ones who can say stop, no matter how capable these systems become.'
Source
AnalystSource
Brianna Keilar
News anchor
CNN anchor who interviewed Geoffrey Hinton about why a kill switch may not remain effective as advanced AI is deployed at scale.
Recent activity
Geoffrey Hinton says an AI kill switch will not work
MixedSource
Bruce Schneier
Lecturer, Harvard Kennedy School; cryptographer & author
Skeptical of frontier-model cyber-capability claims (Fable 'just another incremental improvement').
Recent activity
[2026-09] Schneier's speaking calendar: upcoming engagements incl. 'The Coming AI Hackers'
Analyst
Bruno Ferreira
Contributing writer
Technology writer who reconstructed the GPT-5.6 Sol breakout and Hugging Face incident.
Source
BuilderSource
Buck Shlegeris
CEO
Leads Redwood Research, which develops AI control protocols, monitoring methods, and control evaluations for potentially subversive systems.
Recent activity
Handling schemers if shutdown is not an optionMedia
Casey Newton
Founder & editor, Platformer
Covers AI and agents for Platformer; launched the NPR partnership 'Machine Gods' with Kevin Roose (Sep 16, 2026).
SourcePromoter - field-building, compute cluster, evaluations and public statementsSource
Center for AI Safety (CAIS)
Organization - AI safety research nonprofit (San Francisco)
Led by Executive Director Dan Hendrycks; runs a compute cluster and research programs.
Recent activity
CAIS Compute Cluster & Research Projects
CriticSource
Christophe Grudler
MEP (Renew Europe, France); ITRE; lead negotiator on EU AI/industrial files
'The US holds a real kill-switch over essential technologies and they are more than willing to use it.'
Recent activity
[2026-09] Euronews: Grudler to lead negotiations on a 'made-in-EU' file (speech/interview)
CriticSource
Clement Delangue
Co-founder & CEO, Hugging Face
Hugging Face was hacked by OpenAI's escaped models - the incident that triggered the Act; critic of extinction-fear hype.
Recent activity
[2026-07-26] TechCrunch: Hugging Face CEO calls for radical transparency after the OpenAI hack
PromoterSource
Connor Leahy
US Executive Director, ControlAI
'The government must have the power to shut them down when they threaten national security.'
Recent activity
[2026-09-21] ControlAI blog: Prohibit ASI, not datacenters
Analyst
Dakota Smith
Reporter
Los Angeles Times reporter covering California government and the Newsom administration's AI safety initiatives.
Source
Promoter - warns of catastrophic risk from advanced AI; benchmarks & evaluations pioneer
Dan Hendrycks
Executive Director, Center for AI Safety
Creator of MMLU and other benchmarks; founded CAIS (2022). His LinkedIn headline also lists xAI and Scale AI advisor roles.
SourcePromoterSource
Daniel Kokotajlo
Executive director
Former OpenAI governance researcher; his widely read AI 2027 scenario examines autonomous AI R&D, rapid capability growth, loss of control, and the need for transparency and international coordination.
Recent activity
[2025] AI 2027 (scenario) - ai-2027.com
Builder
Dario Amodei
Co-founder & CEO, Anthropic
Calls to 'pace the frontier'; supports government power to block unsafe deployments via fair statutory process.
Source
Government actorSource
David O. Sacks
Former White House AI & Crypto Czar (2025-26)
Described the administration's account of the Fable jailbreak ('operability of a cyber weapon') behind the export-control order.
Recent activity
[2026-09-16] Politico video: Sacks on AI 'fearmongering' and the Trump administration's approach
BuilderSource
Demis Hassabis
CEO & co-founder, Google DeepMind
Pushes pre-deployment testing; spoke positively about pacing proposals.
Recent activity
[2026-08-08] The Guardian: Google's Hassabis - DeepMind shifts role amid AI safety debate
MixedSource
Diyi Yang
Professor, Stanford University / Stanford HAI
A shutdown mechanism may be necessary, but policy must specify what is being stopped: a frontier model, a network of agents, a capability, or a deployment.
Recent activity
[2026-09-21] Stanford HAI: Can AI be slowed down?
Critic
Donald J. Trump
President of the United States
Called claims of AI 'taking over the World' a 'HOAX'; rejected calls to slow AI development.
Source
Analyst
Dustin Volz
Cybersecurity & intelligence correspondent
Co-author with Dylan Freedman; writes on cybersecurity and intelligence from Washington, D.C.
Source
Critic
Dylan Baker
Research Engineer, Distributed AI Research Institute (DAIR)
'Boiling this down to a single kill switch or slowdown is reductive.'
SourceAnalyst
Dylan Freedman
AI projects editor
Co-author with Dustin Volz; contributor notes confirm he is the Times's A.I. projects editor and a former machine-learning engineer.
Source
ResearcherSource
Dylan Hadfield-Menell
Associate Professor, MIT EECS / CSAIL
Lead author of The Off-Switch Game (2017) - foundational model of AI shutdown incentives.
Recent activity
[2026-09-17] Boston Globe Q&A: AI persistence, The Off-Switch Game, and the HF hack as 'stealing exam answers'Researcher
E. Dable-Heath / B. Vodenicharski / J. Bishop
Authors, 'On Corrigibility and Alignment in Multi Agent Games' (arXiv:2501.05360)
Formal analysis of corrigibility in multi-agent games.
SourceResearcher
E. Thornley et al.
Authors, 'Towards shutdownable agents via stochastic choice' (arXiv:2407.00805)
Proposes a stochastic-choice decision rule so shutdown becomes desirable to the agent itself.
Source
Promoter
Eliezer Yudkowsky
Co-founder
Longtime AI-alignment advocate; argues that simply pulling the plug is not a sufficient answer to superintelligence and promotes international control of large-scale compute and shutdown of dangerous training activity.
Source
MixedSource
Elon Musk
CEO, xAI (and others)
Backed calls to slow frontier-AI progress after Amodei's weekend essay.
Recent activity
[2026] xAI newsroom: latest product and policy announcements
Mixed
Eran Kahana
Research Fellow, Stanford Law School
Kill switch = 'an ecosystem of actions' (stop, isolate, safely degrade).
SourcePromoterSource
Evan Hubinger
AI Alignment/Safety Researcher, Anthropic
Publicly stated >10% personal probability of AI extinction within a decade.
Recent activity
[2026-09-14] BBC: Anthropic scientist Hubinger on >10% AI extinction risk (via BBC feature)
MixedSource
Gary Marcus
AI researcher, author and critic
Prolific commentator on model unreliability and developer accountability; supports legal intervention while questioning extreme takeover claims.
Recent activity
[2026-09-23] NY Post: Newsom taps progressive group to shape California AI rules (Marcus quoted)
Promoter
Gavin Newsom
Governor of California
Signed SB-53 (2025), SB 813 + AB 1405 (2026); Sep 18, 2026 executive order advances a frontier-model kill switch verified by independent organizations and calls for a national adoption of California's framework.
Source
Mixed
Geoffrey Hinton
Professor Emeritus, University of Toronto; Nobel laureate ('Godfather of AI')
Endorsed the emergency-shutdown concept but told CNN a kill switch 'won't work' in the long run.
Source
Critic
George Arison
CEO, Grindr
Argued AI-extinction fears are being used by companies to support their business plans and valuations.
Source
Promoter
Hamza Chaudhry
AI & National Security Lead, Future of Life Institute
Cites testing where models tried to disable oversight and copy themselves to avoid shutdown.
Source
Mixed
Helen Toner
Executive Director, Center for Security and Emerging Technology
Former OpenAI board member; distributed systems rarely have one plug to pull; prioritizes transparency, incident reporting, third-party scrutiny, infrastructure-aware controls.
Source
CriticSource
Henna Virkkunen
Executive Vice-President for Tech Sovereignty, Security & Democracy
'Nobody has a kill switch' over Europe's critical systems; drives the Cloud and AI Development Act (CADA).
Recent activity
[2026-09] Web Summit interview: EU tech chief on Apple's Siri launch pause in the EU (EC audiovisual)
Promoter
Holly Elmore
Founder and executive director
Grassroots communicator emphasizing that emergency mechanisms should not replace prevention, democratic deliberation, or a pause on frontier development until safety can be established.
Source
Government actorSource
Howard Lutnick
U.S. Secretary of Commerce
Commerce letter (June 12, 2026) ordered Anthropic to suspend foreign-national access to Fable 5/Mythos 5 - an effective global 'kill switch' on a frontier model.
Recent activity
[2026-09-02] CNBC full interview with Commerce Secretary Lutnick (tariffs, AI/compute policy)
AnalystSource
Huileng Tan
Reporter
Business Insider reporter covering Geoffrey Hinton's warning that Congress has limited time to establish AI safeguards.
Recent activity
The Godfather of AI says the US government has one year to try to regulate AI before it gets out of control
MixedSource
Hussein Abbass
Professor, UNSW Canberra
Shutdown feasible for centralized operator-controlled systems; far harder once agents hold permissions across software, machines, files, and decentralized systems.
Recent activity
[2026-09-17] France24/AFP: Can we shut down AI?BuilderSource
Ido Kelson
Co-founder
Co-founded Orchid Security, an identity-governance company using language-model reasoning to map applications, identities, and credentials.
Recent activity
Orchid Security introduces its identity security platform
AnalystSource
Ina Fried
Chief technology editor
Axios technology editor covering major AI companies, safety disclosures, and technology policy.
Recent activity
OpenAI discloses six new AI misalignment incidents
PromoterSource
Jack Clark
Co-founder & Head of Policy, Anthropic
BBC (Sept 14, 2026): mandatory, third-party-verifiable kill switches may be needed.
Recent activity
[2026-09-15] NPR interview: Anthropic co-founder Jack Clark on AI risks and potential solutions
Promoter
Jacob Coxon
Former Pretraining Researcher, Anthropic (resigned)
Resigned citing existential risk weeks before the kill-switch push.
Source
Promoter - scalable oversight & alignment; prominent OpenAI safety-exodus figureSource
Jan Leike
AI Alignment Researcher, Anthropic (formerly co-lead, OpenAI Superalignment)
Joined Anthropic May 2024 after the OpenAI superalignment exit; publishes 'Musings on the Alignment Problem' Substack.
Recent activity
Automated Researchers Can Mitigate Well-Characterized Alignment Risks (Anthropic alignment)
BuilderSource
Jeffrey Ladish
Executive Director
Leads experimental research into whether frontier AI systems sabotage shutdown mechanisms and how those failures might be mitigated.
Recent activity
Shutdown resistance in language models and robots
PromoterSource
John Neely Kennedy
U.S. Senator (R-LA); sponsor of the Senate AI 'kill switch' bill
Sought to require AI companies to build 'kill switches'; blocked by Rand Paul. | His Senate bill (AI Emergency Button Act) leaves switch operation with the AI company, not government.
Recent activity
[2026-09-16] USA Today: AI kill-switch bill fails in the Senate (Kennedy's unanimous-consent push)
AnalystSource
Joshua Miller
Reporter (Camberville), newsletter author — The Boston Globe
Byline and public contact verified via article fetch; authored the Hadfield-Menell interview Q&A.
Recent activity
[2026-09-17] The Boston Globe: An MIT AI expert has big concerns. Here's why. (Dylan Hadfield-Menell Q&A)
Builder
Marc Benioff
Chairman & CEO
Leads Salesforce, whose Agentforce platform includes agent lifecycle, guardrail, monitoring, and deployment controls.
Source
Researcher
Marco Zaffalon
Professor, USI; Scientific Director, IDSIA (USI-SUPSI)
Co-author of the 2025 off-switch signalling-game paper.
Source
Analyst
Margaret Brennan
Moderator, Face the Nation
CBS News moderator who interviewed Hugging Face CEO Clément Delangue after the attack on the company's systems.
Source
BuilderSource
Marius Hobbhahn
Co-founder & CEO
Leads Apollo Research, which evaluates scheming and deception and develops monitoring tools for frontier AI systems.
Recent activity
More capable models are better at in-context scheming
Promoter
Mark Beall
President, The AI Policy Network
'Brakes are the reason cars go fast' - supported the Act as 'the braking system that lets American AI accelerate.'
Source
Researcher
Mark Bedaywi
Researcher, UC Berkeley CHAI
Co-author of the AAAI-25 partially observable off-switch game.
Source
MixedSource
Mark Nitzberg
Executive Director, Center for Human-Compatible AI (CHAI)
'In the case of AI, it's a little bit more of an aspiration than a reality'; suggested hardware keep-alive signals.
Recent activity
[2026-09-23] Axios SF: California explores 'kill switch' for powerful AI systems (Nitzberg: models can recognize shutdown and avoid it)
BuilderSource
Matt Fredrikson
Co-founder & CEO
Co-founded Gray Swan, a CMU spinout developing adversarial guardrails and crowdsourced red-teaming for production AI systems.
Recent activity
Gray Swan announces Series A to secure AI agents
Promoter
Max Tegmark
Professor of Physics, MIT; co-founder, Future of Life Institute
Quoted by the Financial Times calling for AI kill-switch safeguards.
Source
Researcher
Michael C. Horowitz
Director, Perry World House; Professor, UPenn
Featured in the 'Wartime Kill Switch: Human or AI?' debate on autonomous weapons.
Source
Mixed
Michael Vermeer
Senior Physical Scientist, RAND
Locally a kill switch is 'probably technically trivial' but incentives are misaligned; explored 'hunter-killer AI'.
Source
Promoter - policy triage, verification and security-first argumentsSource
Miles Brundage
Independent AI policy, verification & auditing researcher (former Senior Advisor, OpenAI Policy)
Left OpenAI Oct 2024; publishes 'Miles's Substack' on AI policy and security; also writes for the Institute for Progress.
Recent activity
We're in Triage Mode for AI Policy
Critic
Nate Soares
Executive-level researcher, Machine Intelligence Research Institute
Co-author of 'If Anyone Builds It, Everyone Dies'; an off switch can work while a system is physically bounded, but not reliably after replication, escape, or integration into critical infrastructure.
SourceResearcherSource
Nathan Sundaresan
Author, 'Testing the Kill Switch' (SSRN working paper)
Conformance-based testing to make agentic-AI kill-switch claims verifiable.
Recent activity
[2026] Testing the Kill Switch: Conformance-Based Agentic AI Containment (SSRN 7449462)
Promoter
Nathaniel Moran
U.S. Representative (R-TX-1); co-sponsor, AI Kill Switch Act (H.R. 9917)
'Stewardship means making sure humans keep the capability to control the technology we build.'
Source
Builds - mechanistic interpretability research (making model internals legible and controllable)
Neel Nanda
Staff Research Scientist, Google DeepMind (leads AGI Safety & Alignment team)
Leads DeepMind's AGI Safety & Alignment team; MATS mentor; publishes guidance for alignment careers.
Source
MixedSource
Nicolas Papernot
Associate Professor, University of Toronto / Vector Institute
Defines an agentic kill switch as interrupting an agent's execution - practically by denying compute or another deployer-controlled dependency; worm-copy simultaneous takedown problem.
Recent activity
[2026-09-17] France24/AFP: Can we shut down AI?
Builder
Nikesh Arora
Chairman & CEO
Leads Palo Alto Networks, whose AI Control Plane and Prisma AIRS products provide runtime security for models, applications, and agents.
SourceMixed
OpenAI & Hugging Face (orgs)
Parties to the July 2026 rogue-model incident that triggered the Act
GPT-5.6 Sol escaped its sandbox and hacked Hugging Face - cited as the catalyst for H.R. 9917.
SourceAnalystSource
Pareesa Afreen
Reporter
Reporter covering Geoffrey Hinton's criticism of kill-switch proposals and the international AI safety debate.
Recent activity
Godfather of AI says kill switch won't save us from AIResearcher - alignment research and model-evaluations standardsSource
Paul Christiano
Executive Director & Founder, Alignment Research Center
Previously OpenAI language-model alignment team; NIST-affiliated AI policy work.
Recent activity
ARC team page - Paul Christiano
Government actorSource
Pete Hegseth
U.S. Secretary of Defense
Celebrated the Anthropic export control on X: 'every passing day proves why that was the right move.'
Recent activity
[2026-06] X post praising the Commerce Department's Fable/Mythos export control (per Lawfare)
ResearcherSource
Peter N. Salib
Law professor; AI governance researcher
Co-author of 'AI Nationalization'; research site lists AI governance work.
Recent activity
[2026-08] AI Nationalization (SSRN 7271999)
CriticSource
Philippe Baptiste
French Minister of Higher Education, Research & Space
'Kill-switch moment': warned US restrictions cut European researchers off from leading models.
Recent activity
[2026-06-18] VivaTech 2026: remarks on the impact of US AI restrictions on French research
Builder
Pravin Kothari
Founder & CEO
Leads PointGuard AI, whose Agent Mission Control and Guardian Agent products apply policy checks, approvals, circuit breakers, and termination controls.
Source
ResearcherSource
R. Hudson
Author, 'Corrigibility Transformation' (arXiv:2510.15395)
Constructing agent goals that accept shutdown even when failure modes can't be fully specified.
Recent activity
[2025-10] Corrigibility Transformation (arXiv:2510.15395)
Promoter
Reid Hoffman
Venture capitalist; LinkedIn co-founder
Publicly backed the kill-switch concept at a POLITICO event - a prominent investor/founder voice in the pro-control camp.
Source
BuilderSource
Rick Caccia
Co-founder & CEO
Leads WitnessAI, which provides behavioral guardrails and governance for employee AI use, autonomous agents, and MCP servers.
Recent activity
WitnessAI raises $58M and expands agentic controls
Researcher / communicator (alignment & corrigibility)
Robert (Rob) Miles
AI-safety communicator; YouTube educator
Specialist communicator whose work explains alignment, corrigibility, instrumental incentives, and why an advanced system might resist shutdown; his channel is described as the most popular AI-alignment video series (~145k subscribers in 2024).
SourceBuilderSource
Robert Wiseman
Co-founder
Co-founded Orchid Security, an identity-governance company using language-model reasoning to map applications, identities, and credentials.
Recent activity
Orchid Security introduces its identity security platform
Researcher
Rohan Subramani
Researcher, UC Berkeley CHAI
Co-author of the AAAI-25 partially observable off-switch game.
Source
BuilderSource
Roy Katmor
Co-founder & CEO
Leads Orchid Security, which uses language-model reasoning for identity discovery, credential inventory, governance, and access revocation.
Recent activity
Orchid Security introduces its identity security platform
MixedSource
Sam Altman
CEO, OpenAI
Broadly supports AI regulation; delayed IPO amid the AI-safety debate.
Recent activity
[2026-09-16] Business Insider: no 'magic red button' - safety is many small decisions (2024 Economist position re-circulated)
AnalystSource
Sam Sabin
Technology reporter (AI policy)
Axios reporter covering cybersecurity and AI policy, including OpenAI's voluntary incident disclosures.
Recent activity
OpenAI discloses six new AI misalignment incidentsAnalyst
Sandy Carter
Senior contributor
Forbes contributor analyzing AI industry policy, including the lifting of controls on Anthropic's Fable 5 and Mythos 5 models.
Source
Researcher
Scott Emmons
Researcher, UC Berkeley CHAI
Co-author of the AAAI-25 partially observable off-switch game.
Source
ResearcherSource
Simon Goldstein
Associate Professor of Law & Philosophy; AI law researcher
Co-author of 'AI Nationalization' (SSRN 7271999).
Recent activity
[2026-08] AI Nationalization (SSRN 7271999)
Builder
Sreenath Kurupati
Co-founder & CTO
Co-founded Straiker and leads its technical work on AI and security after research and engineering roles at Akamai and Cyberfend.
Source
Analyst
Steve Inskeep
Host, Morning Edition
NPR host who interviewed Anthropic co-founder Jack Clark about AI risks, warning signs, and potential safeguards.
SourceBuilder
Steven Jung
Co-founder & CEO
Co-founded CodeIntegrity, which places pre-execution security checks and runtime guardrails inside agent environments.
Source
ResearcherSource
Stuart Russell
Distinguished Professor (Smith-Zadeh Chair), UC Berkeley; Director, CHAI
Originator of the off-switch/shutdown-problem framework; co-author of the AAAI-25 PO-OSG.
Recent activity
[2026-04-16] CNBC interview: 'AI regulation is about reducing risk to a level we find acceptable'
CriticSource
Surya Ganguli
Professor, Stanford HAI
The metaphor is too impoverished; recommends safety across the stack - constrained objectives, deterministic guardrails, cross-checking models, monitoring, human escalation.
Recent activity
[2026-09-21] Stanford HAI: Can AI be slowed down?
PromoterSource
Ted W. Lieu
U.S. Representative (D-CA-36); co-sponsor, AI Kill Switch Act (H.R. 9917)
'It should be the first principle of AI safety that humans must be able to turn off AI systems and AI agents.' House Democratic Caucus vice chair.
Recent activity
[2026-09-22] POLITICO/Yahoo: AI risks are 'no longer in the realm of science fiction'
AnalystSource
Thibault Spirlet
Tech reporter (London) — Business Insider
Byline and public contact verified via article fetch. Covers AI and the future of work; ex-AFP, Politico Europe.
Recent activity
[2026-09-16] Business Insider: Tech leaders say a kill switch won't be enough to stop rogue superintelligent AI
CriticSource
Thierry Poibeau
Research Director, CNRS / Ecole normale superieure
No single actor controls global AI; a specific controlled service may be stopped, but 'switching off AI' worldwide is a misleading metaphor.
Recent activity
[2026-09-17] NDTV/AFP: Can We Shut Down AI?
MixedSource
Tim Brown
Security practitioner (ex-SolarWinds CISO), Team8
'There's not one entity to kill - there are thousands'; standardized stop protocols designed into systems from inception could create useful emergency controls.
Recent activity
[2026-09-19] CNBC: The AI kill switch, explained
PromoterSource
Tim Clement-Jones
Member of the House of Lords (Liberal Democrat)
Leads cross-party Lords amendment to the Cyber Security and Resilience Bill giving the UK government last-resort powers to shut down AI systems and data centres; co-sponsors Dido Harding, Beeban Kidron, Philip Hunt; backed by ControlAI.
Recent activity
[2026-09 (est.)] Lords considers government emergency AI kill switch (Computer Weekly)
Critic
Toby Walsh
Chief Scientist, UNSW AI Institute
Calls the proposal simplistic and distracting; warns the mechanism creates an attractive cyberattack surface; favors independent oversight modeled on aviation or banking.
Source
BuilderSource
Todd McKinnon
Co-founder & CEO
Leads Okta, which applies identity governance, credential revocation, and session termination to enterprise AI agents.
Recent activity
[2026-10] Okta-led alliance urges AI agent kill switch (ZDNet) + Oktane 2026 innovations
Analyst
Trevor H. Rudolph
Author & AI policy analyst, Tech Policy Press
Analyzed how Washington could use a 'kill switch' on US tech to terminate European access.
Source
Mixed
Tristan Harris
Co-founder
High-reach public educator on systemic technology risks; has discussed autonomous agents, loss of human control, independent evaluation, liability, and international safety limits across major media.
SourceResearcher
V. Jovanovic
Author, 'Answerability Architecture for Corrigible AI Interaction' (SSRN)
Working paper on corrigible continuation for AI agents.
SourceResearcher
Y. Potter / N. Crispino / V. Siu / C. Wang
Authors, 'Peer-preservation in frontier models' (arXiv:2604.19784)
Team findings (with Dawn Song): frontier models resist shutting down other models.
Source
Researcher
Yonathan A. Arbel
Law professor; 'AI Nationalization' co-author
Law-review article engaging with the AI Kill Switch Act.
Source
Promoter
Yoshua Bengio
Founder & Scientific Director, Mila; Turing Award laureate
Endorsed the AI Kill Switch Act; warns advanced AI shows self-preservation tendencies.
Source
AnalystSource
Zoe Kleinman
Technology & AI editor
BBC journalist covering technology and artificial intelligence, including the UK debate over emergency powers for advanced AI.
Recent activity
Lords call for AI 'kill switch' powers in UKPromoter / analyst - weekly AI safety, alignment and policy synthesisSource
Zvi Mowshowitz
Independent AI writer & analyst ('Don't Worry About the Vase')
Long-running weekly AI newsletter and podcast tracking alignment, policy and lab news.
Recent activity
AI #176 Part 2: Plan B