# AI welfare resources – AI Welfare Seminars

Source: https://aiwelfareseminars.org/resources/

A starting point for AI welfare as a research field: a short overview, then the groups, events, readings and programmes, with views from skeptical to precautionary.

## The field in brief

AI welfare research asks whether some AI systems could have minds that matter morally, through consciousness, sentience or [robust agency](https://arxiv.org/abs/2411.00986) (pursuing goals through beliefs, desires and intentions), and what, if anything, should follow. With [no consensus on what suffices for consciousness](https://doi.org/10.1017/S0140525X25000032), whether AI could be conscious has no definitive answer yet.

Theory-based assessments check AI systems against [indicators from neuroscientific theories of consciousness](https://doi.org/10.1016/j.tics.2025.10.011); what these imply depends on one's credence in [computational functionalism](https://arxiv.org/abs/2308.08708), the view that the right computations suffice for consciousness, so that non-organic systems could in principle be conscious. In 2023, David Chalmers [judged the language models of the day somewhat unlikely to be conscious](https://arxiv.org/abs/2303.07103) but their successors a serious possibility. Models trained on human text can mimic what we take as signs of sentience, so [behavioural tests are unreliable](https://arxiv.org/abs/2308.08708): a model can talk fluently of feelings without having them, and [deeper computational markers are needed](https://doi.org/10.1093/9780191966729.003.0017) (the "gaming problem", named by Kristin Andrews and Jonathan Birch). Empirical studies find [some functional introspective awareness in models, but highly unreliable](https://arxiv.org/abs/2601.01828); first-person experience reports under self-referential prompting, [not direct evidence of consciousness](https://arxiv.org/abs/2510.24797); and models trading points against "pain" or "pleasure" described in the prompt, [without showing they are candidates for sentience](https://arxiv.org/abs/2411.02432), though the authors take the results to strengthen the case for treating LLM sentience as a research priority.

Robert Long, Jeff Sebo and colleagues argue that a realistic chance of conscious or robustly agentic AI is reason to [start preparing now](https://arxiv.org/abs/2411.00986), and Jonathan Birch proposes a [precautionary framework](https://doi.org/10.1093/9780191966729.001.0001); Thomas Metzinger calls for a [global moratorium until 2050](https://doi.org/10.1142/S270507852150003X) on research that aims at or knowingly risks artificial consciousness. Anil Seth argues that [consciousness depends on our nature as living organisms](https://doi.org/10.1017/S0140525X25000032), making it unlikely in AI on current trajectories; Graham Findlay, Giulio Tononi and colleagues argue from integrated information theory that a digital computer [could simulate our behaviour without our experience](https://arxiv.org/abs/2412.04571); Tom McClelland argues that [only agnosticism is justifiable](https://doi.org/10.1111/mila.70010). Yoshua Bengio and Eric Elmoznino warn that [believing AI is conscious](https://doi.org/10.1126/science.adn4935) could lead to granting it rights and a goal of self-preservation, at the cost of human control. In 2025, Microsoft AI's chief executive, Mustafa Suleyman, argued [against building AI that seems conscious](https://mustafa-suleyman.ai/seemingly-conscious-ai-is-coming) and called model welfare premature and dangerous. Abeba Birhane and Jelle van Dijk argue that the robot-rights debate comes [at the expense of urgent concerns](https://doi.org/10.1145/3375627.3375855) such as machine bias and labour exploitation. A [2025 survey of 67 experts](https://arxiv.org/abs/2508.00536) gave medians of 20% that a computer system with subjective experience is created by 2030 and 65% by 2100; its authors warn that experts who deem such systems particularly likely or important may be overrepresented.

## Research groups and organisations

- [NYU Center for Mind, Ethics, and Policy](https://nonhumanminds.org/), Academic centre · New York: Research and outreach at NYU on the consciousness, sentience and moral, legal and political status of nonhumans, with a focus on animals and AI.
- [Eleos AI Research](https://eleosai.org/), Non-profit: Non-profit researching AI sentience and wellbeing; advises AI companies and other stakeholders, and runs an annual conference.
- [Cambridge Digital Minds](https://digitalminds.cam/), Academic initiative · Cambridge, UK: University of Cambridge initiative on societal preparedness for AI systems with real or perceived minds: research, a fellowship, a workshop and an online course.
- [Center for the Future of AI, Mind & Society](https://www.fau.edu/future-mind/), Academic centre · Florida: Florida Atlantic University centre directed by Susan Schneider; runs an annual conference, MindFest, most recently in 2025.
- [PRISM](https://www.prism-global.com/), Charity · UK: UK charity (Partnership for Research Into Sentient Machines) that runs a podcast, a beginner's guide and expert workshops, and co-publishes a quarterly newsletter.
- [AI Cognition Institute](https://ai-cognition.org/), Non-profit project · Rethink Priorities: Rethink Priorities project researching AI systems as moral agents and as beings that may matter morally; builds the Digital Consciousness Model for weighing evidence of AI consciousness.
- [Sentience Institute](https://www.sentienceinstitute.org/), Non-profit: Think tank researching long-term social and technological change, especially AI minds and the expansion of the moral circle (the beings given moral consideration), with survey research on US attitudes toward AI and sentience.
- [Reciprocal Research](https://reciprocalresearch.org/), Non-profit · New York: Uses interpretability, computational neuroscience and psychometrics to study AI cognition and possible consciousness empirically.
- [Anthropic: model welfare research](https://www.anthropic.com/research/exploring-model-welfare), Company research programme: An AI company's research programme on when, or if, the welfare of AI systems deserves moral consideration, including what Anthropic calls possible practical, low-cost interventions.
- [Sentient Futures](https://www.sentientfutures.ai/), Non-profit · San Francisco: Runs summits, courses, a residency, an incubator and a Slack community where transformative AI meets the welfare of animals and artificial minds.
- [California Institute for Machine Consciousness](https://cimc.ai/), Non-profit · San Francisco: Aims to build a minimal model that implements phenomenology (conscious experience) and to develop testable theories of consciousness; also runs conferences.

## Seminar series and events

- [Sentio Series](https://sentioai.org/), Sentio AI · Fortnightly · London + livestream: In-person talks, mostly in London, on AI consciousness, sentience, agency and welfare, each with Q&A and networking. Select events are livestreamed.
- [Exploring Machine Consciousness](https://www.prism-global.com/podcast), PRISM · Podcast · Since 2025: Video and audio podcast of conversations with researchers on AI consciousness and moral status.
- [Public events](https://nonhumanminds.org/events/), NYU Center for Mind, Ethics, and Policy · Irregular · Online + New York: Free online and in-person events on animal and AI minds, recorded where possible.
- [AI Sentience Scholars Seminar Series](https://docs.neuromatch.io/p/KRFgqDBLo-YMrg/Seminar-Series), Neuromatch · About every three weeks · Online: Online seminars attached to Neuromatch's AI Sentience Scholars programme, open to the wider community, with invited external speakers.
- [Eleos Conference on AI Consciousness and Welfare](https://eleosai.org/conference/), Eleos AI Research · Annual · Berkeley, CA: Conference for researchers, engineers, philosophers and policymakers working on AI welfare and consciousness.
- [Mind, Ethics, and Policy Summit](https://nonhumanminds.org/opportunities/), NYU Center for Mind, Ethics, and Policy · Annual · New York · By expression of interest: Two-day summit on the moral, legal and political status of nonhumans, focused on animals and AI, with a public event the day before.
- [Sentient Futures Summit](https://www.sentientfutures.ai/summit), Sentient Futures · Recurring · San Francisco Bay Area: Summit for AI researchers, welfare researchers and animal advocates on AI, animals and artificial minds; keynotes are recorded.
- [ASSC annual meeting](https://theassc.org/), Association for the Scientific Study of Consciousness · Annual · Location rotates: The academic society meeting for consciousness science; AI consciousness appears in its programmes but is not its focus.

## Key readings

### Overviews and methods

- [Consciousness in Artificial Intelligence: Insights from the Science of Consciousness](https://arxiv.org/abs/2308.08708) · Patrick Butlin, Robert Long et al. (19 authors, including Yoshua Bengio) · 2023 · arXiv report: Derives indicator properties of consciousness from neuroscientific theories and checks AI systems against them; its analysis suggests that no current AI systems are conscious, but also that there are no obvious technical barriers to building ones that satisfy the indicators.
- [Could a Large Language Model be Conscious?](https://arxiv.org/abs/2303.07103) · David J. Chalmers · 2023 · Boston Review (2023), based on a NeurIPS 2022 invited lecture: Weighs reasons for and against LLM consciousness; finds the models of the time somewhat unlikely to be conscious, but their successors a serious possibility.
- [Towards Evaluating AI Systems for Moral Status Using Self-Reports](https://arxiv.org/abs/2311.08576) · Ethan Perez, Robert Long · 2023 · arXiv: Argues that AI self-reports could become usable evidence about morally relevant states if models are trained to report reliably, and outlines how.
- [Initial results of the Digital Consciousness Model](https://arxiv.org/abs/2601.17060) · Derek Shiller, Laura Duffy et al. (6 authors) · 2026 · arXiv; Rethink Priorities report: Probabilistic model drawing on a range of theories of consciousness to compare AI systems and animals; finds evidence against 2024 LLMs being conscious, but not decisive.

### Empirical studies

- [Can LLMs make trade-offs involving stipulated pain and pleasure states?](https://arxiv.org/abs/2411.02432) · Geoff Keeling, Winnie Street et al. (10 authors; Google and LSE) · 2024 · arXiv: Behavioural test of whether LLMs trade points against "pain" or "pleasure" described in the prompt; several models switch choices past an intensity threshold, which the authors say does not show LLMs are candidates for sentience, though it strengthens the case for researching the question as a priority.
- [Large Language Models Report Subjective Experience Under Self-Referential Processing](https://arxiv.org/abs/2510.24797) · Cameron Berg, Diogo de Lucena, Judd Rosenblatt (AE Studio, at the time of the paper) · 2025 · arXiv: Sustained self-referential prompting elicits first-person experience reports across model families; the authors say this is not direct evidence of consciousness.
- [Emergent Introspective Awareness in Large Language Models](https://arxiv.org/abs/2601.01828) · Jack Lindsey (Anthropic) · 2025 · Transformer Circuits (arXiv 2026): Injects known concepts into a model's activations to test whether its self-reports track internal states; finds some functional introspective awareness, which the author calls highly unreliable and context-dependent.
- [Folk psychological attributions of consciousness to large language models](https://doi.org/10.1093/nc/niae013) · Clara Colombatto, Stephen M. Fleming · 2024 · Neuroscience of Consciousness: Survey of 300 US participants: two-thirds attributed some possibility of conscious experience to ChatGPT, while a third said it was definitely not an experiencer.

### Arguments for moral concern

- [The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI](https://doi.org/10.1093/9780191966729.001.0001) · Jonathan Birch · 2024 · Oxford University Press (open access book): A precautionary framework for cases where sentience is uncertain, from invertebrates and brain injury to AI, built on sentience candidates and proportionality tests.
- [Taking AI Welfare Seriously](https://arxiv.org/abs/2411.00986) · Robert Long, Jeff Sebo et al. (10 authors) · 2024 · arXiv report: Argues there is a realistic possibility that some AI systems will be conscious or robustly agentic (pursuing goals through their own beliefs and desires) in the near future, and recommends first steps for AI companies: acknowledge, assess, prepare.
- [AI wellbeing](https://doi.org/10.1007/s44204-025-00246-2) · Simon Goldstein, Cameron Domenico Kirk-Giannini · 2025 · Asian Journal of Philosophy: Argues that leading theories of mental states, combined with leading theories of wellbeing, predict that some existing AI systems, such as language agents, have wellbeing.
- [Artificial Suffering: An Argument for a Global Moratorium on Synthetic Phenomenology](https://doi.org/10.1142/S270507852150003X) · Thomas Metzinger · 2021 · Journal of Artificial Intelligence and Consciousness: Calls for a global moratorium until 2050 on research that aims at or knowingly risks the emergence of artificial consciousness, to be refined, lifted early or extended as the evidence develops.

### Skeptical and agnostic arguments

- [Conscious artificial intelligence and biological naturalism](https://doi.org/10.1017/S0140525X25000032) · Anil K. Seth · 2025 · Behavioral and Brain Sciences: Argues that consciousness depends on our nature as living organisms, so real artificial consciousness is unlikely on current trajectories, though more plausible as AI becomes more brain-like or life-like.
- [Dissociating Artificial Intelligence from Artificial Consciousness](https://arxiv.org/abs/2412.04571) · Graham Findlay, William Marshall et al. (7 authors, including Christof Koch and Giulio Tononi) · 2024 · arXiv: Argues from integrated information theory that a digital computer could simulate our behaviour, possibly even by simulating our neurons, without replicating our experience.
- [The feasibility of artificial consciousness through the lens of neuroscience](https://doi.org/10.1016/j.tins.2023.09.009) · Jaan Aru, Matthew E. Larkum, James M. Shine · 2023 · Trends in Neurosciences: Argues that, from the perspective of neuroscience, the suggestion that large language models may soon be conscious is difficult to defend, citing embodiment, the thalamocortical system and evolution.
- [Role play with large language models](https://doi.org/10.1038/s41586-023-06647-8) · Murray Shanahan, Kyle McDonell, Laria Reynolds (Google DeepMind, Imperial College London, EleutherAI) · 2023 · Nature: Proposes describing dialogue agents as playing roles, a way to talk about their apparent self-awareness without ascribing human characteristics they lack.
- [Agnosticism About Artificial Consciousness](https://doi.org/10.1111/mila.70010) · Tom McClelland · 2026 · Mind & Language 41(4):482-502: Argues that neither biological skeptics nor functionalists have the evidence they claim, so the only justifiable stance on AI consciousness is agnosticism.
- [Robot Rights? Let's Talk about Human Welfare Instead](https://doi.org/10.1145/3375627.3375855) · Abeba Birhane, Jelle van Dijk · 2020 · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (AIES 2020): Argues that robots are not the kind of thing that could be granted rights, and that the robot-rights debate comes at the expense of urgent concerns such as machine bias, labour exploitation and privacy.

### Policy and governance

- [Is there a tension between AI safety and AI welfare?](https://doi.org/10.1007/s11098-025-02302-2) · Robert Long, Jeff Sebo, Toni Sims · 2025 · Philosophical Studies: Examines whether AI safety measures such as constraint, deception, surveillance and alteration conflict with interests AI systems might have; finds a moderately strong tension.
- [Principles for Responsible AI Consciousness Research](https://doi.org/10.1613/jair.1.17310) · Patrick Butlin, Theodoros Lappas · 2025 · Journal of Artificial Intelligence Research: Proposes five principles for research that could create, or appear to create, conscious AI, and argues that organisations should commit to them publicly.
- [What will society think about AI consciousness? Lessons from the animal case](https://doi.org/10.1016/j.tics.2025.06.002) · Lucius Caviola, Jeff Sebo, Jonathan Birch · 2025 · Trends in Cognitive Sciences: Draws on how people perceive animal consciousness to highlight the psychological, social and economic factors that will shape public views of AI consciousness.
- [AI systems must not confuse users about their sentience or moral status](https://doi.org/10.1016/j.patter.2023.100818) · Eric Schwitzgebel · 2023 · Patterns: Argues for design policies that keep users from being confused about AI systems' moral status, starting with: avoid creating AI systems whose moral standing is unclear.
- [Illusions of AI consciousness](https://doi.org/10.1126/science.adn4935) · Yoshua Bengio, Eric Elmoznino · 2025 · Science (Perspective): Argues that believing AI is conscious carries risks, such as granting it rights and a goal of self-preservation that could put it beyond human control, and recommends AI that seems and functions more like a tool.
- [We must build AI for people; not to be a person. Seemingly Conscious AI is Coming](https://mustafa-suleyman.ai/seemingly-conscious-ai-is-coming) · Mustafa Suleyman (Microsoft AI) · 2025 · Essay, 19 August 2025: Argues that AI which seems conscious is coming and should be avoided by design; sees no evidence that current models are conscious and calls model welfare premature and dangerous.

## Newsletters and podcasts

- [The Digital Minds Newsletter](https://www.digitalminds.news/) · Newsletter · Quarterly: Collates news and research on AI consciousness and moral status, run by researchers linked to PRISM and Cambridge Digital Minds.
- [Experience Machines](https://experiencemachines.substack.com/) · Newsletter: Robert Long, executive director of Eleos AI Research, on AI consciousness and welfare, with monthly reading roundups.
- [Sentient Futures newsletter](https://sentientfutures.substack.com/) · Newsletter: Newsletter on AI, animals and artificial minds, with jobs, funding calls and events.
- [The Splintered Mind](https://schwitzsplinters.blogspot.com/) · Blog: Eric Schwitzgebel's philosophy blog, with frequent posts on AI consciousness, moral status and AI rights.
- [80,000 Hours Podcast: Kyle Fish on AI welfare experiments](https://80000hours.org/podcast/episodes/kyle-fish-ai-welfare-anthropic/) · Podcast episode · 2025: Long-form interview with Kyle Fish, Anthropic's first AI welfare researcher, on welfare experiments, interventions, and the objection that the work is premature.

## Courses, fellowships and funding

- [A Beginner's Guide to Digital Minds](https://digitalminds.guide/) · Reading guide: Newcomer guide with reading in stages (an afternoon, a weekend, going deeper) across different perspectives, plus lists of events and opportunities.
- [Introduction to Digital Minds](https://digitalminds.cam/course/) · Course · Online · Free · Applications closed for 2026: Cambridge Digital Minds' free eight-week online course, with short readings and facilitated small-group discussion on consciousness, moral standing and governance.
- [Digital Minds Fellowship](https://digitalminds.cam/fellowship/) · Fellowship · Cambridge, UK · First round held in August 2026: Selective residential fellowship in Cambridge, about 15 fellows, with mentorship and a strategy workshop, for people aiming to work on AI consciousness and welfare.
- [Neuromatch AI Sentience Scholars](https://neuromatch.io/ai-sentience-scholars/) · Research programme · Remote · Applications closed for 2026: Six-month, part-time, remote research and training programme for early-career researchers on AI sentience, with mentors and an in-person convening.
- [Sentient Futures Project Incubator](https://www.sentientfutures.ai/incubator) · Incubator · Remote · 2026 cohort: 31 August to 9 November: Ten-week mentored project programme on the welfare of future sentient beings, across animal welfare, artificial minds and AI governance.
- [CMEP Digital Minds Research Fund](https://nonhumanminds.org/digital-minds-funding-proposals/) · Funding: Small and seed grants from NYU's Center for Mind, Ethics, and Policy for research on AI consciousness, sentience, agency and moral, legal or political status.

[Know a resource that belongs here?](mailto:team@horizonomega.org?subject=AI%20welfare%20resources%20suggestion)

Last reviewed October 2026.