Last time we looked at what happens when AI starts handling one of the most crucial jobs in any company: deciding who keeps their job and who loses it. We covered that lawsuit against Meta, where 26 former employees claimed the company used AI systems to help identify workers for redundancy, which included people who had taken maternity leave or disability leave.
The case has now reached a new stage: if AI helps make employment decisions, who exactly is responsible when workers believe those decisions were unfair, especially if people who aren’t in the company can see how the tech was used?
What Has Happened Since The Lawsuit Was Filed?
A US judge has refused to temporarily stop Meta from completing the redundancies, according to Reuters. The ruling does not settle the case but, it does show just how difficult it can be for employees to prove an employer relied on AI when making redundancy decisions.
US District Judge William Orrick said the workers faced a basic obstacle because, as he says, “they were not in the rooms where it happened.”
The employees claim AI systems influenced redundancy decisions, but they cannot easily prove how those systems were used because Meta controls almost all of the information. Judge Orrick also said he had to accept Meta’s explanation for now because the workers had not produced evidence to challenge it.
The judge has not closed the case and he will later decide whether to issue a preliminary injunction that could return the employees to their jobs until their individual legal cases finish. He also said he could reach a different decision if the workers produce evidence “regarding whether and how AI was used in an improper manner.” A hearing is scheduled for 24 August.
Why Is Proving An AI Case So Tricky?
According to Reuters, legal specialists say this lawsuit helps explain why many people expected a flood of AI employment cases, but very few have reached court.
One reason is that workers often have little knowledge of how AI systems operate within their workplace. Many employees have also signed arbitration agreements. These agreements usually stop workers from bringing class action lawsuits or presenting their cases before a jury, sending disputes into private arbitration instead.
Christine Webber, co chair of the civil rights and employment practice at Cohen Milstein Sellers & Toll, said confidential arbitration can stop information becoming public.
She said, “Even if you establish that a particular system would produce discriminatory outcomes left and right, you have no way of sharing that information with other employees.”
Reuters reported the Meta workers are also bound by arbitration agreements. Each employee must usually pursue an individual case rather than joining together in a public courtroom.
Who Is Responsible When AI Helps Make Workplace Decisions?
The workers claim Meta consulted internal AI assisted systems when selecting jobs for redundancy. According to the lawsuit, these systems tracked productivity, AI token usage, workplace communications and activity, including a large language model assistant known as Metamate and productivity scoring built from keystrokes, screen content, emails and browser history.
Meta rejects those claims and court filings and company statements say people made every decision connected to the nearly 8,000 redundancies announced earlier this year. A Meta spokesperson also said the company had no further comment after last week’s ruling.
The legal case now comes down to an issue many workplaces could face. If software helps rank employees, but managers approve the final decision, who is responsible when workers believe the outcome was unfair?
The court has not answered it, so now, it’ll be a matter of waiting to see if the employees can uncover enough evidence to show exactly how AI was used behind closed doors.
We’ve asked experts who they think should be responsible in such cases, and here’s what they shared…
Our Experts:
- Rob McKellar, Legal Services Director, Peninsula
- Catherine Mitchell, Employment Partner, HCR Law
- Mahendra Balal, Founder, Sovereix
- Juan Mathews Rebello Santos, Cybersecurity Researcher & Ethical Hacker, Founder, BNVD.org
Rob McKellar, Legal Services Director, Peninsula
“The current legal system is not set up to cater with AI making dismissal decisions which means that dismissals should remain under human control. Employers remain ultimately responsible for decisions made under their authority, even if they rely on AI systems.
“Employer actions, including those that may result in unfair or wrongful dismissals, whether by AI or human decision, can lead to legal claim. Employers are expected to provide fair procedures, reasonable notice, and lawful reasons for dismissals. If an AI system fails to meet these standards, the employer could be held liable for the consequences, and an employment tribunal judge is not likely to accept a defence of “It was AI’s decision to dismiss”.
“Because of this, it is essential that humans continue to make dismissal decisions, even if AI has been used during the process, to ensure that decisions remain fair and reasonable, free from bias or discrimination that may be present in AI decisions.”
More from News
- The Paramount-Warner Bros Merger Pause Exposes Streaming’s Growing Identity Crisis
- How Did A Rogue OpenAI Model Manage To Hack Its Rival Company?
- The DBIST Has Business, Innovation, Science And Trade…But Where’s The Technology?
- TikTok Exploited By Industrial-Scale Counterfeit Networks Targeting Luxury Brands
- What Is Samsung’s “Galaxy” Credit Card, And Who Can Use It?
- HireQuotient Extends AI Recruiting Capabilities To Paylocity Customers In Frontline Industries
- Tech Stocks Are Crashing Today – Here’s Why
- Could Space Stocks Become “The Next AI”?
Catherine Mitchell, Employment Partner, HCR Law
“When large organisations try to outsource the decision-making to AI they are tangoing with risk.
“Employment law holds employers responsible for decisions affecting employees. If an employer uses AI they must understand that it is a tool and should never be allowed to make a decision autonomously.
“AI won’t understand the nuances of an employee’s role. And it won’t be able to run a fair process without a human steering it. An organisation can feed all the information it has into AI and allow it to run algorithms but we are already seeing large organisations get it wrong. There is no nuance, no human understanding of the issue. As a result, an employer could find themselves facing significant claims from those who feel they have been wrongfully or unfairly dismissed.
“Humans should always be at the centre of the decision-making process, especially in employment and definitely when dismissing workers. These are real people’s lives and real businesses footing the bill for getting it wrong.”
Juan Mathews Rebello Santos, Cybersecurity Researcher & Ethical Hacker, Founder, BNVD.org
“Accountability starts with whoever deployed the system without a human-in-the-loop review mechanism. The vendor built the model, but the employer chose to replace a human decision with a black box output. In practice, liability falls on the employer because they control the termination process and are legally the decision maker. But the real gap is that most AI hiring and firing tools ship with disclaimers that shift risk entirely onto the customer while the vendor keeps no liability for outcomes.
“When a model hallucinates a performance metric or misweights a tenure factor, no one owns that error contractually. At BNVD.org we have analysed termination disputes where employers could not explain which features drove the AI decision because the model was a third party SaaS API with no audit trail. The fix is regulatory: require employers to run a parallel human review for any AI driven termination and mandate that vendors expose the feature weights and confidence intervals of each decision. If the vendor cannot explain the output, they should not be indemnified.”
Mahendra Balal, Founder, Sovereix
“When an AI makes an error that results in the termination of a worker, that is on the enterprise’s executive leadership and the human resources management team, not the software vendor and certainly not the algorithm itself.
“As an analyst at Sovereix, I focus on AI integration and enterprise risk, and I see AI in terms of capital expenditure and tooling. AI is a sophisticated operational tool, not a scapegoat. If a physical machine fails in a factory and injures a worker, the management is liable both legally and operationally for not having proper safety procedures and supervision in place.
“Algorithmic management is subject to the same unit economics of liability. A company that deploys an autonomous AI agent for important HR decisions without a human-in-the-loop failsafe or proper audits for algorithmic drift puts efficiency before risk management. You can ask an AI to do the work of performance evaluation, but you can’t outsource the fiduciary and legal responsibility for the result.”
