AI at work: Fair, accurate or biased? Data centres are getting hot. The UK in search of AI regulation.

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 “🔒”.

Misaligned Recap

New in Misaligned this week is “The Erasure of Interaction”, in which Ioannis Akingonte looks at the consequences of AI “note taking”. Also, to fit in with the topic of this week’s newsletter, we want to point to our report from the University of Greenwich workshop on “The Ethics of AI in Workplaces”.

British bits

The UK Competition and Markets Authority (CMA) has imposed a new conduct requirement for Google search, which must take specific action under the UK’s digital markets competition regime. Google is now required to make sure that publisher content is properly attributed, using clear links in AI‑generated search results (➚UK Government).

Meanwhile, the Metropolitan police in London have been granted a 12-month extension to a pilot project that the police have been running with controversial US-firm Palantir, while the force will carry out a procurement process. That comes only weeks after the mayor of London, Sadiq Khan, blocked a £50m deal between the Met and the US company (➚The Guardian).

US bits

OpenAI has reportedly discussed giving a 5% stake to the US government, amounting to almost $50 billion. The AI company seeks to “clear political obstacles” by securing financial buy-in from the Trump administration. The talks are reportedly “conceptual” and are in the early stages (➚🔒Financial Times).

Regulatory bits

A senior official at the Bank of England has called for an updated regulatory framework to cater for the inexorable rise of agentic AI: “Our frameworks were not built to contemplate autonomous ⁠agents, ‌and relying on a human in the loop for all agent actions is unlikely to be realistic,” said deputy governor Sarah Breeden (➚Finextra). Only last week the head of the Financial Conduct Authority complained that AI is “outrunning regulation”, which is not that hard if you do not have such regulation in place.

Singapore’s Personal Data Protection Commission (“PDPC”) has issued proposed guidelines (➚Public Consultation) on the use of personal data in generative artificial intelligence systems (➚Hogan Lovells). The public consultation that ran for one month is now closed.

Data Centres

While the heat wave is still in full swing, a community living next to the largest datacentre park in Europe in Slough, UK, says the scorching summer heat has become unbearable (➚The Guardian). A recently published paper suggests datacentres create a heat island effect, and states “that the data heat island effect could have a remarkable influence on communities and regional welfare in the future, hence becoming part of the conversation around environmentally sustainable AI worldwide” (➚Arxiv).

The Scottish government could be about to consider a sweeping moratorium on building new datacentres. The Scottish National party (SNP)’s national council passed a motion to freeze all new datacentres in Scotland. That motion has been sent to the Scottish government for consideration. It would apply to all datacentre projects that have not yet received planning permission (➚The Guardian).

Meanwhile, an investigation by The Guardian has revealed that “OpenAI does not appear to have visited one of Stargate UK’s key sites — and that £20bn of the “potential” £30bn in investment touted by the UK government appears to have been totally hypothetical” (➚The Guardian).

At university

Two-in-five UK universities have no AI policy that a student, parent or regulator can easily find online, according to What UK University AI Policies Actually Do: A Study of 96 Institutions (➚Hepi). Key findings of the study involve that 41% of UK degree-awarding institutions have no publicly accessible AI policy, and the dominant model is conditional trust: “students are trusted only if they declare their use, retain evidence and submit to verification.”

At work

Workday must face claims that its popular AI-powered human resources software weeded out job applicants at other companies in ways that violated California law and a federal ​ban on discrimination against workers with disabilities, a federal judge ruled on Monday (➚Reuters).

The judge wrote that because Workday allegedly participated in unlawful conduct from its ​California headquarters, it could be held liable for discrimination under state law. Workday says it provides more than 11,000 organizations with AI-powered cloud solutions, including human capital management and recruiting products used by employers around the world.

The case against Workday is one of the most significant cases over bias in recruitment. Plaintiffs allege that Workday’s “smart” tools disparately impacted them by deprioritizing applicants based on protected characteristics as well as automatically incorporating previous employer bias (➚Employment Law Worldview).

We tackled the ethical issues of AI in recruitment in article last year.

Science bits

This week’s paper of interest is “Is It Fair to be Accurate? Moral-Emotional Responses to Organizations’ AI Orientation Choices” by Vancompernolle Vromman et al.

The paper looks at how the discrepancy between a fair and an accurate algorithm is perceived and hypothesizes “that an organization’s choice between an algorithm maximizing accuracy at the expense of fairness and one prioritizing fairness over accuracy triggers distinct moral-emotional responses among third-party observers”.

The paper attempts to advance “management research on algorithmic decision-making and extends deonance theory to algorithmic human resource management, establishing AI orientation choices as a moral context informing observers’ approval or disapproval of organizations.”

In conclusion the paper finds that “this research demonstrates that people, as third-party observers, do not interpret organizations’ algorithmic choices as morally neutral”.

Vancompernolle Vromman, F., Hericher, C., Vande Kerckhove, C., & Raineri, N. (2026). Is It Fair to be Accurate? Moral-Emotional Responses to Organizations’ AI Orientation Choices. Business & Society, 43.

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.