Employees at Palantir and Google wonder who are the baddies, techno fascism, apologies and critical looks at LLMs in health care.

Welcome to a new edition of misaligned bits, the (roughly weekly) newsletter from Misaligned where we sum up recent news and research, sometimes with a lighter touch.

As usual, we will mark all non-medium links with “➚” (external link) and all possibly paywalled links with “🔒”.

Palantir related bits

Minneapolis campaigners are pressing the Swiss National Bank to dump its investment in Palantir: A delegation from Minneapolis travelled to the bank’s shareholder meeting in Bern last Friday to present the city council’s request (➚Reuters). Meanwhile, current and former employees of Palantir are reportedly expressing dismay about the company’s “mini manifesto” (➚Wired). Wired also reports that Palantir management has been holding a handful of “ask me anything” forums across the company.

Talking about Palantir, its founder Peter Thiel is reportedly pouring millions into the new start-up “Objection” that seeks to “rate” journalists with the help of a “jury of AI models”. Especially it appears to give journalists a “low rating” if they rely on anonymous sources, potentially chilling whistleblowers. (➚TechCrunch)

The Baddies

While the Anthropic lawsuit against the Pentagon is dragging on, Google has told its employees it “proudly” works with the US military and will continue to do so. This comes as Google has reportedly faced backlash from employees after it signed a deal with the defence department that will allow its AI technology to be used in classified operations (➚🔒Financial Times). The contract appears to mirror those of OpenAI and xAI have signed with the DoD.

Science terminated

Multiple scientists who served on National Science Board established to guide the US’s nearly $9 billion basic science funding agency were terminated from their positions by President Trump last Friday (➚🔒Washington Post). It is unclear how many members of the board were dismissed and if the Trump administration plans to replace them.

Data Centres moratoriums

The Seminole Nation of Oklahoma has become the first Indigenous nation to officially ban data centre construction from lands under its jurisdiction: (➚Native News Online) On March 7, the Nation’s Tribal Council approved a resolution by a 24–0 vote to “implement a moratorium on the advancement of generative Artificial Intelligence (AI) technology and hyperscale data center development within the Seminole Nation and within tribal lands and territories.” The resolution was proposed after a startup approached the Tribal Council, asking them first to sign a non-disclosure agreement (NDA) and a letter of intention to develop a data centre on Seminole lands.

Apologies

Sam Altman of OpenAI has written a letter apologizing that his company didn’t alert law enforcement about the online behaviour of a person who shot and killed eight people in Tumbler Ridge, British Columbia (➚AP News).

This comes after a judge in another case ordered that OpenAI, its CEO Sam Altman must defend a federal court lawsuit alleging a man’s ongoing interactions with the artificial intelligence platform led him to kill his mother and himself (➚Case 3:25-cv-11037-RS).

Legal bits

The United States acting attorney general wants to intervene in the xAI lawsuit against Colorado’s AI regulation (➚Case 1:26-cv-01515-DDD-CYC). Following this, Colorado’s AI antidiscrimination law, which was set to take effect June 30, won’t be enforced until the federal court rules on a forthcoming motion from xAI to block it. (➚Bloomberg Law)

Science Bits

This week we have three studies we found noteworthy:

The first paper this week is not the usual study, but a paper by Mark Coeckelbergh published in Springer Nature looking at the recent political developments coming out of Silicon Valley: “Drawing from the classic philosophical and political theory literature on fascism, authoritarianism, and totalitarianism, this paper argues that and how features of contemporary digital technologies, their governance, and their political context mirror features of fascism.”

The author comes to the conclusion that “Technofascism is not a replication of twentieth-century fascism but an adaptation” and notes that “recognizing these parallels and origins is crucial for developing strategies to resist new forms of domination that threaten democratic forms of life in the digital age.”

Coeckelbergh, M. Technofascism: AI, Big Tech, and the new authoritarianism. AI & Soc (2026).

A conference paper published last year looks at the role LLMs play in mental health counselling. The study notes that LLMs “were not designed to replace healthcare workers, but they are being used in ways that can lead users to overestimate the types of roles that these systems can assume.” The paper also notes that the anthropomorphic responses of chatbots can “create a false sense of emotional connection” and generally create a discriminatory environment where “individuals who are ‘knowledgeable enough’ to correct LLM outputs are at an advantage, while others, due to lack of clinical knowledge and digital literacy, are more likely to suffer from clinically inappropriate responses”.

Iftikhar, Z., Xiao, A., Ransom, S., Huang, J., & Suresh, H. (2025). How LLM Counselors Violate Ethical Standards in Mental Health Practice: A Practitioner-Informed Framework. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 8(2), 1311–1323.

A study by Arya S Rao et al asks the question if off-the-shelf large language models (LLMs) can demonstrate reliable performance across the clinical workflow: “In this cross-sectional study of 21 LLMs, frontier LLMs achieved high accuracy on final diagnoses but performed poorly in generating differential diagnoses and navigating uncertainty relative to other reasoning stages.”

It concludes that “off-the-shelf LLMs have not yet achieved the intelligence required for safe deployment and remain limited in demonstrating advanced clinical reasoning”.

Rao AS, Esmail KP, Lee RS, et al. Large Language Model Performance and Clinical Reasoning Tasks. JAMA Netw Open. 2026; 9(4):e264003.

Follow and subscribe

If you stumbled across this article on the web, subscribe to the “misaligned bits” newsletter and follow Misaligned on Medium, LinkedIn, Threads or Mastodon.

You can now also subscribe to this newsletter directly with your email.