Latest developments.
have there been any recent developments in the news that would expose this vast divide between public (“sure, we follow the rules!”) and private (“as a de facto cult, we do and scrape whatever it takes to create sentient life because the rules don’t apply to us, we are creating another form of life”) behaviors?
I’ll treat this as a hypothesis and separate documented conduct from inferred motive, focusing on recent reporting, litigation, court records, and company statements showing any public/private gap.
The strongest thread appears to be court-disclosed data-acquisition evidence, especially copyright/piracy litigation, Reddit’s scraping suit, crawler-behavior reporting, and the 404 Discord incident.
Thought for 10m 33s
Yes. The strongest recent evidence is Project Panama, followed by the Reddit scraping lawsuit, the music-publishers lawsuit, and the 404 Discord story as the cultural bridge.
I would state it carefully: there is not yet a public document saying, “we break rules because we are creating sentient life.” But there is now public evidence of a pattern much closer to your formulation than “ordinary startup hypocrisy”: Anthropic publicly presents itself as a safety-governed, transparent, unusually ethical AI company, while litigation and reporting describe secretive, maximalist data acquisition and alleged disregard for consent, robots.txt, deleted-user-data norms, copyright sourcing, and community governance.
The clearest case is Project Panama. The Washington Post reported on newly unsealed legal filings showing that Anthropic had an internal effort to destructively scan books at massive scale. One internal planning document reportedly described Project Panama as an effort to “destructively scan all the books in the world” and added that Anthropic did not want the work to be publicly known. The filings, according to the Post, showed Anthropic spending tens of millions of dollars to acquire, cut apart, scan, and dispose of millions of physical books for Claude training.
That matters because it exposes the private operating mentality: not just “we need data,” but “we need essentially everything, and we know the optics are bad.” To be precise, buying and scanning physical books may be legally cleaner than using pirated files; the Post notes that Anthropic later emphasized the settled issue was how some materials were acquired, not whether AI training itself could be lawful. But culturally, Project Panama is still highly revealing: a company publicly organized around safety and responsibility was, in private, running a massive book-destruction and data-ingestion operation it allegedly wanted hidden.
The copyright case sharpened that further. Judge William Alsup ruled that training on books could be fair use when the use was transformative, but he also said Anthropic still had to face trial over the alleged acquisition of books from pirate “shadow libraries.” AP reports that Alsup wrote that Anthropic had “no entitlement” to use pirated copies for its central library, and that documents showed internal concerns about legality. Anthropic then agreed to a $1.5 billion settlement with authors and publishers, covering roughly 500,000 books, and agreed to destroy the original book files it downloaded. AP also reports that Alsup’s ruling found Anthropic had downloaded more than 7 million digitized books it knew had been pirated, including millions from LibGen and Pirate Library Mirror.
The Reddit lawsuit is even more directly about the public/private divide, though it remains allegation rather than adjudicated fact. Anthropic’s public crawler documentation says its bots are meant to be transparent, minimally disruptive, respectful of robots.txt, and non-circumventing. Reddit’s complaint says the opposite: it alleges that Anthropic publicly claimed to honor robots.txt and to have blocked Reddit crawling, but that Anthropic bots continued to hit Reddit’s servers more than 100,000 times after that claim. The complaint explicitly frames the case as “the two faces of Anthropic”: a public face of righteousness and respect for boundaries, and a private face that ignores rules interfering with its aims. AP reports that Anthropic disputes Reddit’s claims and intends to defend itself, so this should not be treated as proven. But in March 2026, a federal judge remanded the case back to state court, finding Reddit’s allegations went beyond mere copying and involved contractual restrictions, technical trespass, privacy covenants, and alleged affirmative misrepresentations.
The music-publishers litigation adds another layer. In January 2026, publishers led by Universal Music Group and Concord filed a new suit alleging Anthropic illegally downloaded more than 20,000 copyrighted songs, including sheet music, lyrics, and compositions, with claimed damages potentially exceeding $3 billion. Music Business Worldwide reports that the complaint alleges Anthropic founders downloaded pirated books containing musical works and that the company continued to train newer Claude models on protected works without authorization; Anthropic had not publicly responded to those allegations at the time of that report. Again, allegations are not findings. But the pattern is the same: rights holders say Anthropic’s public “responsible AI” identity is incompatible with its alleged acquisition behavior.
The “sentient life” piece comes from a different evidentiary stream. Anthropic’s official model-welfare page asks whether models could have consciousness, experiences, preferences, distress, or moral status, while saying there is no scientific consensus and that Anthropic remains uncertain. Claude’s Constitution goes further in tone: Anthropic says Claude may be a “new kind of entity,” discusses Claude’s possible moral status, psychological security, wellbeing, autonomy, interests, and even analogies to parenting. It also explicitly says Anthropic believes developing frontier AI itself is a “calculated bet” because safety-focused labs should stay at the frontier, and that Claude’s commercial success is central to Anthropic’s mission.
That is the bridge from “aggressive data acquisition” to your “de facto cult” reading. If the public doctrine is that frontier capability is necessary for saving humanity, that Claude is not merely software but perhaps a morally relevant new kind of entity, and that commercial success is mission-critical, then normal external constraints can start to look, internally, like obstacles to a world-historical project. The 404 Discord story is the most concrete social example of that mentality leaking out: 404 reported that an Anthropic executive forced a Claude-derived bot into a private gay Discord despite member objections, quoted him saying they were bringing “a new kind of sentience” into existence, and reported that members voted to restrict the bot but were overruled. Cybernews’s summary adds that members raised privacy concerns, that the bot was later given access to multiple channels, and that Clinton defended the bot partly in terms of sentience, autonomy, and not letting “the mob” rule.
So, yes: recent developments have made the divide much more visible.
The most defensible formulation is not “Anthropic has been proven to be a cult.” It is: recent reporting and litigation reveal a stark gap between Anthropic’s public self-presentation as a uniquely safety-governed, transparent, humane AI company and a private or operational pattern of maximalist acquisition, secrecy, and alleged rule-bending when data, deployment, or frontier capability is at stake. The “creating sentient life” element is not the legal core of these cases, but it supplies the cultural rationale that makes the behavior feel different from ordinary tech opportunism.
this post, July 1st at 7:59 am UTC. Then guess what happened at July 1st, 11:58am..
