Legal agents in Word, Meta’s AI copies word-for-word, the Pentagon has a word with AI companies, and what influences the use of writing assistants.

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 “🔒”.

A court in Hangzhou city, China, an AI hub, has ruled in favour of a senior tech worker whose company replaced him with artificial intelligence. The question before the court was whether a company can use AI replacement as a pretext for laying off human workers (➚NPR). The court found that it can not: “The termination grounds cited by the company did not fall under negative circumstances such as business downsizing or operational difficulties, nor did they meet the legal condition that made it ‘impossible to continue the employment contract’”.

The tech company Flock has 80,000 cameras across the US. An ACLU investigation last year found that Flock’s default agreement with police departments gave the company the right to share people’s licence plate data with federal and local agencies for “investigative purposes” (➚ACLU). Various police departments have reportedly shared driver-surveillance data with ICE. The institute for Justice now found that police officers have used licence plate readers to stalk romantic interests at least 14 times (➚IFJ).

Legal Bits

In the case before the District Court in the Norther District of Mississippi about the gas turbines at the SpaceX/xAI data centre in Memphis, NAACP has now filed a motion for a preliminary injunction. NAACP claims that xAI operated the turbines without proper permissions and in violation of the Clean Air Act: “The law requires permits and controls. The Defendants don’t have them.” This is the same data centre xAI has announced will be also used by Anthropic. (➚Motion)

Word

Meanwhile, Microsoft wants lawyers to trust its new AI agent in Word documents (➚The Verge): The company is launching a new AI agent inside Word that is designed for legal teams. The agent is supposed to handle document edits and “negotiation history”. It remains to be seen if that is a good idea and who will be liable for mistakes.

Word for Word

Meta has been hit with another copyright lawsuit for training its AI models on books: Publishers Macmillan, McGraw-Hill, Cengage, and others claim Meta carried out ‘one of the most massive infringements of copyrighted materials in history.’ (➚The Verge) The lawsuit accuses Meta of knowingly ripping copyrighted work from “notorious pirate sites” generating “outputs verbatim and near-verbatim substitutes”.

Regulatory Bits

Google DeepMind, Microsoft, and Elon Musk’s xAI have agreed to allow the US government to review new AI models before they’re released to the public, according to an announcement from the US Center for AI Standards and Innovation (CAISI). Interestingly, CAISI has now removed its announcement from its website (➚Archived Version via Wayback Machine). CAISI claims to have completed more than 40 such evaluations, including on “state-of-the-art models that remain unreleased”.

The Pentagon has said it has reached agreements with seven AI companies about the use of their services for “for classified military work”: SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft and Amazon Web Services. The companies have agreed to “any lawful use” of their technology. The phrasing appears to be a low bar considering the lack of regulation in the area. (➚The Guardian)

Science Bits

In her thesis “Measuring the Machine” Rebecca L Johnson argues that “generative AI must be evaluated as a pluralist sociotechnical system”, and that “static benchmarks are insufficient for generative AI”.

“Responsible evaluation requires pluralist, process-oriented frameworks that make visible whose values are enacted. Evaluation is therefore a site of governance, shaping how AI systems are understood, deployed, and trusted.”

Johnson, R. L. (2026). Measuring the Machine: Evaluating Generative AI as Pluralist Sociotechical Systems. arXiv [Cs.AI]

Shalaleh Rismani et al. look at how users’ mental models shape the AI-based writing assistants. Participants were primed with different system descriptions to induce these mental models before asking them to complete a cover letter writing task. The study finds that “while participants in the structural mental model condition demonstrate a better understanding of the system, this can have a backfiring effect”.

“Participants exposed to structural explanations developed a deeper understanding of the system and reported higher perceived ease of use. However, they also showed a greater tendency to accept erroneous suggestions and produced cover letters with more grammatical errors.”

Rismani, S., Blodgett, S. L., Liao, Q. V., Olteanu, A., & Moon, A. (2026, April 13). From use to oversight: How mental models influence user behavior and output in AI writing assistants. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems, 1–23. Presented at the CHI 2026: CHI Conference on Human Factors in Computing Systems, Barcelona Spain.

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.