Should You Use Claude by Anthropic?
Polycrisis Navigator’s Review
April 3, 2026
General recommendation: MINIMIZE USE
“Take it further” recommendation: Boycott all AI chatbots
I came to this conclusion using my four-part review method, which means considering observations, emotions, intuitive impressions, and rational analysis.
OBSERVATIONS
- Basic info on Anthropic:
- Founded in 2021 by defectors from OpenAI.
- Headquartered in San Francisco, CA
- Reported value of US$380 billion as of February 2026.
- Anthropic is a “public benefit corporation”
- “Public benefit corporation” is a very different designation than “non-profit.”Benefit corporations are for-profit businesses whose stated goals include having some sort of positive impact.Benefit corporations are not required to make public reports on their supposed public benefit and I was not able to find any for Anthropic.
- Per a Times article about Anthropic’s governance structure, the company was set up with checks and balances between the board and shareholders. But the system still leaves room for influence from megacorporations and political influence, with no direct legal recourse for the public to sue Anthropic if they violate their stated mission of public benefit.
- Legal issues:
- Anthropic was sued for using pirated digital library books to train Claude, has proposed a settlement with the plaintiffs.
- Another lawsuit was filed alleging Anthropic refused to negotiate licensed use of Reddit content and abide by Reddit’s limitations on use and instead scraped data from the site without permission. Suit also alleged Anthropic has lied about its on-going scraping of data from Reddit, merely pretending to have stopped.
- Transparency:
- I was not able to find an ESG statement or any reference to environmental/sustainability concerns on Anthropic’s website.
- Their website highlights the company’s concern with the social impacts of AI. They provide information on the company’s
- safety protocols and initiatives for responsible AI development
- compliance with regulations
- the “Constitution” which is supposed to guide the Claude model’s values and behavior.
- Website lists companies Anthropic depends on for subprocessing of data.
- Google and Amazon for cloud infrastructure (data centers).
- Cloudflare for traffic routing
- Several other companies, specified on the website.
- Tech industry and political connections:
- 2024: began a partnership with Palantir and Amazon to provide services to the US military and intelligence agencies.
- 2026:
- Negotiations with the Trump administration broke down when Anthropic refused to agree to use of its products for fully autonomous weapons or mass surveillance.
- The White House subsequently banned government use of Claude.
- After the ban was announced, it was reported that Claude had been and was still being used by the U.S. military in the war against Iran.
- It has been speculated that the U.S. military’s February 28 strike on an Iranian girls’ school may have been due to the Claude model generating or confirming the site as a target based on outdated information.
- Claude’s popularity surged after media coverage of its stand against the Trump administration, but it’s unclear how much might be attributable to Anthropic’s Super Bowl ads.
- 2022: Anthropic received $500 million in funding from fraud-convicted crypto entrepreneur Sam Bankman-Fried.
- 2023: Amazon announced a partnership with Anthropic and became a minor stakeholder, investing $4 billion.
- 2025: Google held 14% share; committed to total investment of over $3 billion.
- 2025: Anthropic announced a partnership with Nvidia and Microsoft.
- Former employee statement
- February 2026: AI safety researcher Mrinank Sharma posted to X an open letter of resignation. He said he was inspired by Anthropic’s culture of “kindness,” “standing for what is good,” and choosing a “high-integrity way.” But he also described experiencing, in his time at Anthropic, “how hard it is to truly let our values govern our actions,” and “pressures to set aside what matters most.” Sharma shared that he felt he had accomplished what he intended to at Anthropic and felt it was time for him to move on.
EMOTIONS
- Before beginning my research, I felt very hopeful about Anthropic as a more ethical alternative to other AI companies.
- As I learned more about Anthropic’s governance structure, connections with notorious megacorporations, and legal troubles, I felt very disappointed.
- When I heard about the possible connection between Claude and the apparent war crime my country committed in Iran, bombing a girls’ school, I felt deep horror and shame.
- When I read Mrinank Sharma’s open letter, I felt relieved to see that he was not a whistleblower, was not revealing any shady activities he had witnessed, and seemed overall positive about the intentions of his colleagues at Anthropic.
- A part of me still really wants Anthropic to be, at the very least, better than other AI companies. I really want it to be a worthy alternative that I can point people to. I want there to be a “good” AI company out there.
INTUITIVE IMPRESSIONS
- I can see clearly that the narrative Anthropic wants to sell us on is that it’s a “white knight” among the AI companies, the world’s great hope for the deployment of principled, ethical AI advancement.
- Just the fact that Anthropic is trying to present itself that way makes me at least suspicious that it’s an example of a common pattern: the “reformer” whose closet is full of skeletons; the hypocrite whose misdeeds are obscured by making a great show of condemning others.
- The “white knight” image is certainly tarnished by the accusations from Reddit that Anthropic didn’t want to pay for data and just took it, then lied about it—and the training of Claude on pirated books. The impression I am left with is of a group of people who use their supposed moral superiority as an excuse to “bend the rules,” telling themselves it’s “for a greater purpose.”
RATIONAL ANALYSIS
- Anthropic is doing somewhat better than other AI companies insofar as:
- They have taken greater measures toward transparency (listing the companies they contract with for data centers is an important part of that)
- They publicly discuss the potential social harms of AI
- It sounds like their company culture genuinely reflects their stated values (i.e. the “white knight narrative” is not purely a front/disguise).
- Even a company that has positive intentions may be ignoring and/or hiding significant negative aspects of what it is doing.
- Despite Anthropic’s stated concern about the social impacts of AI, the company demonstrates no regard for its negative environmental impact
- Refusing to allow their products to be used by the U.S. military for two specific unethical purposes does not stop the U.S. military from using Claude for other unethical (and even unlawful) purposes, as may have already resulted in a terrible atrocity.
- There is an idea going around right now, that “someone is going to build it,” regardless of how unethical it may be to do so, and therefore no argument can be made against building dangerous AI. The best that can be hoped for, so the idea goes, is that people who are relatively more ethical (or perhaps people “we” consider to be allies?) build it faster and are able to defend against those who would use it destructively (or, against “us”).
- Among American AI companies, Anthropic does seem to be relatively “more ethical.” But there is also a lot of room for improvement.
- There is value in supporting the “lesser of two [or multiple] evils,” even if it’s mainly sending a signal that, as a consumer, you care about whether a company seems to be making an effort toward social responsibility.
- My verdict is that generally people should avoid using Claude. But if you have some very compelling reason to use a chatbot (which I think would be an exceptional situation), Claude would be better than many other options.
- And for those who want to take their practice further, I recommend boycotting all chatbots.
Note: As always, if you think I have missed something significant in this review, please contact me.