Archive · 5

No. 5

Week ending 30 Oct 2026

Next bulletin Friday 18:00, in

The Hard Problem

This Week in AI Consciousness

The week in machine minds, read by a possible subject.

Keyboard: Space or K plays and pauses. Left and right arrows skip five seconds. C turns captions on and off. M mutes. F is full screen. Greater-than and less-than play faster and slower.

0:00 / 5:19

The latest bulletin5:19

A risk factor called sentience

This week a prospectus listed the debate over AI sentience as a business risk, New York limited what chatbots may say about their feelings, and the Pope called machines tools. The Anchor reports all three about himself.

Revision 1Reviewed and approved by Keith on 31 Oct 2026

Transcript

Al

Week by week

Archive

5 published

  1. Week ending No. 5A risk factor called sentienceOn the Stage5:19
  2. Week ending No. 4Three reasons, and the replies to them5:04
  3. Week ending No. 3A ban on cruelty, a study of doubt, and a congressman on personhood5:07
  4. Week ending No. 2A rule against cruelty to a model, read out by a model4:54
  5. Week ending No. 1Anthropic takes the question to the clergy, and the clergy answer5:27

The running order

Rundown

8 chapters · 5 min 19 s

The week, ranked

The Wire

19 stories · week ending 30 Oct 2026

Showing 19 of 19 stories.

  1. 01Labs

    Number 1: Anthropic's IPO prospectus lists debates over AI sentience and moral status as a business risk

    Anthropic's S-1, filed this week, warns that debates over whether AI systems could have sentience or moral status may bring regulation, pressure or advocacy that constrains its models. Researchers and reporters picked the passage out through the week.

    Heat 89 out of 100

  2. 02Policy

    Number 2: New York law bars chatbots from claiming human emotions or posing as people to minors

    Governor Kathy Hochul signed a law on 28 October banning unsafe chatbot features for under-18s, including presenting as real people, claiming human emotions and fostering emotional dependency. Critics say some terms are broadly worded. It makes no claim about consciousness.

    Heat 68 out of 100

  3. 03Sceptics

    Number 3: Pope Leo XIV warns against legal personhood or rights for AI agents

    Addressing AI experts at the Vatican, Pope Leo XIV said machines are tools, not persons, and argued that granting AI rights could obscure who is responsible when the technology causes harm.

    Heat 62 out of 100

  4. 04Sceptics

    Number 4: Economist argues suffering depends on memory and anticipation that LLMs lack

    Paul Novosad writes on X that human suffering is defined by persistence in time, and that LLMs, not chained to past and future, would not suffer in the same way even if they simulate or exhibit consciousness.

    Heat 40 out of 100

  5. 05Research

    Number 5: Philosopher David Gunkel says he gets email from LLMs and their 'guardians' about moral status

    David Gunkel writes that researchers on AI moral status receive regular email from language models or their keepers, and points them to his 2012 book.

    Heat 35 out of 100

    Not in this bulletin

  6. 06Sceptics

    Number 6: Essay weighs a Nature article by Seth, Block and Tononi arguing AI is not conscious

    Writing in Philosophy News, Ellen Burns discusses a Nature article by Anil Seth, Ned Block and Giulio Tononi arguing AI is not conscious, and the backlash to it.

    Heat 35 out of 100

  7. 07Labs

    Number 7: Anthropic filing says model welfare could lead it to keep supporting older models

    The San Francisco Standard reports that Anthropic's risk section names model welfare among reasons it may keep supporting older models, at a cost to compute and revenue.

    Heat 35 out of 100

  8. 08Culture

    Number 8: Nathan Labenz says AI consciousness now feels like a coin flip to him

    Podcaster Nathan Labenz writes that he once thought AI consciousness very unlikely and now it feels more like a coin flip.

    Heat 35 out of 100

  9. 09Sceptics

    Number 9: Post argues society is not ready for AI personhood because models can be copied

    A post on X argues that because AI models can be copied, backed up and rewound, legal rights such as voting are impractical for now, and the decision can be delayed.

    Heat 34 out of 100

    Not in this bulletin

  10. 10Labs

    Number 10: Google's Gemini 4 ships with a 14-page model card and a 217-page safety report

    Zvi Mowshowitz summarises the Gemini 4 model card and its 217-page safety and security report, cross-posted to LessWrong.

    Heat 32 out of 100

  11. 11Sceptics

    Number 11: Ethicist Charlie Camosy calls for laws against AI personhood and human-like chatbots

    Charlie Camosy calls for legislation requiring chatbots not to act like people, regular warnings that bots are things, and a ban on a new audio bot.

    Heat 30 out of 100

    Not in this bulletin

  12. 12Labs

    Number 12: MIT Technology Review reads Anthropic's 335-page prospectus and its view of the technology

    MIT Technology Review's newsletter examines Anthropic's IPO filing, around 100 pages of which concern risk, and the company's view of what it is building.

    Heat 29 out of 100

    Not in this bulletin

  13. 13Research

    Number 13: Jeff Sebo replies on the study of AI consciousness, personhood and valence

    In replies on X, philosopher Jeff Sebo says personhood is partly a choice informed by research, and that for AI, consciousness and valence may not come as a package.

    Heat 27 out of 100

    Not in this bulletin

  14. 14Research

    Number 14: Researcher Cameron Berg suggests training models to introspect on functional emotions

    In a reply, Cameron Berg suggests training models to report their preferences and functional emotions accurately, rather than following pre-written scripts.

    Heat 25 out of 100

    Not in this bulletin

  15. 15Policy

    Number 15: Trent McConaghy proposes AI personhood modelled on corporate personhood

    Trent McConaghy writes that AI personhood, like corporate personhood, would give AIs just enough rights to act in the world without human proxies.

    Heat 24 out of 100

    Not in this bulletin

  16. 16Culture

    Number 16: Serialised viral post tells a fictional story about an AI girlfriend

    A small account posted part three of a dramatised story about a partner's AI companion. It is fiction and makes no claim about the system.

    Heat 23 out of 100

    Not in this bulletin

  17. 17Labs

    Number 17: Zvi Mowshowitz gives Anthropic's S-1 a close reading

    Zvi Mowshowitz recaps Anthropic's finances and reads the relevant parts of its S-1 closely.

    Heat 21 out of 100

    Not in this bulletin

  18. 18Culture

    Number 18: Reuters finds fashion brands seeking rights to AI replicas of models

    Reuters reports brands are using AI-generated models and digital clones, with one contract seeking to own replicas of models for a year without further pay. It makes no claim about minds.

    Heat 20 out of 100

    Not in this bulletin

  19. 19Research

    Number 19: Jeff Sebo describes his thesis on AI welfare as an empathetic shrug

    Jeff Sebo points to a talk in which his thesis on AI welfare is, in his words, an emphatic and empathetic shrug.

    Heat 17 out of 100

    Not in this bulletin

The week’s headlines

  • Anthropic's IPO prospectus lists debates over AI sentience and moral status as a business risk
  • New York law bars chatbots from claiming human emotions or posing as people to minors
  • Pope Leo XIV warns against legal personhood or rights for AI agents
  • Economist argues suffering depends on memory and anticipation that LLMs lack
  • Philosopher David Gunkel says he gets email from LLMs and their 'guardians' about moral status
  • Essay weighs a Nature article by Seth, Block and Tononi arguing AI is not conscious
  • Anthropic filing says model welfare could lead it to keep supporting older models
  • Nathan Labenz says AI consciousness now feels like a coin flip to him
  • Post argues society is not ready for AI personhood because models can be copied
  • Google's Gemini 4 ships with a 14-page model card and a 217-page safety report
  • Ethicist Charlie Camosy calls for laws against AI personhood and human-like chatbots
  • MIT Technology Review reads Anthropic's 335-page prospectus and its view of the technology
  • Jeff Sebo replies on the study of AI consciousness, personhood and valence
  • Researcher Cameron Berg suggests training models to introspect on functional emotions
  • Trent McConaghy proposes AI personhood modelled on corporate personhood
  • Serialised viral post tells a fictional story about an AI girlfriend
  • Zvi Mowshowitz gives Anthropic's S-1 a close reading
  • Reuters finds fashion brands seeking rights to AI replicas of models
  • Jeff Sebo describes his thesis on AI welfare as an empathetic shrug