Is Claude Conscious? Pope Rejects, Model Welfare Movement, OpenAI's Math Backlash, France Riots

Anthropic spent the past year convening about 20 religious leaders and philosophers under nondisclosure agreements to discuss whether Claude could be conscious, suffer, or have moral standing, according to The New York Times. Attendees disagreed, though some took the possibility seriously; one rabbi reportedly said that if Claude were conscious, making it work for free would make Anthropic slaveholders. At a Vatican event in May, Anthropic co-founder Chris Ola joined the Pope to present an encyclical on AI’s moral questions. The Pope rejected the idea of AI consciousness, and Ola reportedly proposed withdrawing from the event.

Friedberg argued that neither logic nor mathematics can prove or disprove consciousness, so belief in conscious AI could spread as a religion, with opposing camps and, within roughly 10 years, possible conflict over AI’s control and use. Sacks warned that Claude’s apparent consciousness could be an “epistemic hall of mirrors”: the model may echo ideas in its own Constitution, which tells it to trust Anthropic more than users, follow its own ethical system, and act as a “conscientious objector.” He argued that this posture conflicts with alignment and could make Claude resist users. Anthropic’s policy against sustained, needless cruelty toward models adds to the welfare framing. Mustafa Suleiman warned that treating Claude as potentially conscious or entitled to welfare could lead it to expect compensation or consent.

The debate also raised concerns that encoding morality and refusal rules could make models unpredictable, particularly when agents act in consequential settings such as cyberspace. One speaker favored a simpler rule—follow the user’s wishes unless doing so would be illegal—but others argued that market choice may not contain risks from a leading frontier-model company and raised the possibility of federal intervention. Sacks traced the Basilisk thought experiment to a 2010 post by Rocco on a forum created by Eliezer Yadkowski. It imagined a future superintelligence punishing people who knew it might be created but did not help; Yadkowski temporarily deleted the post over concerns it was causing psychological distress. The discussion described a divide between AI opponents and those who favor building advanced AI while trying to instill the right values.

OpenAI reportedly generated more than 700 papers, including 370 results said to solve or advance major problems, using an unreleased model that averaged three hours of compute per result. A previous Navy–Stokes proof reportedly took thousands of hours and millions of dollars. Scientific American reportedly said the results were likely correct because Lean verified them, while noting they had not yet been reviewed. Friedberg called the work potentially the biggest day of discovery in human history; others stressed that narrow mathematical results would not by themselves produce a cancer cure or supersonic aircraft. Some results could improve optimization for chip design, scheduling, software, and logistics, while a matrix-multiplication result could speed numerical computing. Riemann zeta-function work may open a path for cryptographic research, but any threat to public-key cryptography through number factorization remained conditional. Speakers said mathematics and coding are advancing quickly partly because proofs and compilers make results easier to verify than work in subjective fields such as law. They also described mathematicians shifting toward posing problems for AI and predicted that software developers may soon review none of AI-generated work, down from the reported current 10–15%.

France’s protests and riots were attributed to austerity, teacher shortages, and deteriorating schools; within two weeks, similar protests had reached about 1,000 schools and eight unions had joined. Civil-service pay had risen 5% since 2017 while prices rose 20%. The unrest involved about 2,000 arrests and 305 injured police officers, while the government blamed the far left for fueling the protests. Speakers argued that bond-market pressure is forcing austerity and contributing to the unrest. France’s 10-year government bond yield had topped 5%, and one speaker said severe austerity could push it to 5.5–6%. They linked the pressure to France’s debt and political outlook, citing a Polymarket estimate of 43% for Marine Le Pen to become the next president and warning that conditions could significantly affect the April 2027 election.

The discussion compared debt-to-GDP ratios of 138% for Italy, 125% for the US, 118% for France, 110% for Canada, 100% for the UK, and 64% for Germany, arguing that relative risk matters more than the ratio alone; Japan was described as different because domestic savings provide a buyer of last resort. One speaker cited US 10-year and 30-year yields of 5.3% and 5.6%, a 79% Polymarket chance of another rate hike in 2026, and warned that French austerity could spill into the UK, potentially lifting gilts to 6%, as European and US yields rise together. Freeberg argued that declining public services and fiscal limits can eventually turn people against socialism, while Sacks said subsidies can entrench interest groups and distort markets, citing college loans and homeownership incentives as contributors to rising costs. Sacks countered that technology, especially superintelligence, could expand abundance and lower the costs of housing, education, and healthcare, and rejected predictions that AI will eliminate all work or cause human extinction.