An internal Meta post announcing a tracking system used to improve AI assistants received hundreds of angry comments and emojis from workers who feared they were being replaced.
The company launched the software last month, which allowed AI to track each employee’s clicks and keystrokes as they go about their workday.
‘For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples,’ the announcement read, per Business Insider.
More than 100 angry and surprised emojis riddled the comment section as employees voiced their concerns, Reuters reported.
Many workers expressed concern that they were inadvertently educating the AI that might eventually replace them.
‘This makes me super uncomfortable,’ an engineering manager wrote in a comment reviewed by the New York Times. ‘How do we opt out?’
However, Meta’s Chief Technology Officer, Andrew Bosworth, responded that employees would not be able to opt out of the program on their company laptops.
‘Your callousness to the concerns of your own employees is concerning,’ an anonymous employee fired back at Bosworth, according to the Times.
Others were concerned that the ongoing gathering of employee information and behaviours might pose a security risk.
‘This data is very tightly controlled,’ Bosworth said. ‘This will not be a leak risk.’
Meta spokesperson Tracy Clayton told the Daily Mail that the new employee tracking program was only meant to improve AI products.
‘There are safeguards in place to protect sensitive content, and the data is not used for any other purpose,’ he said.
In recent years, Mark Zuckerberg has invested billions in the developing AI sector, pushing the use of AI throughout Meta’s social platforms.
‘I think we know that AI is one of the most competitive fields, probably in history,’ he has said.
The company’s 78,000 workers have been urged to embrace constantly evolving AI policies.
Although Zuckerberg claimed that the software was not being used for ‘surveillance or performance tracking or anything like that.
To offset AI spending, the company has allegedly already planned to slash its workforce by ten per cent, according to an April 17 announcement.
Meta’s Head of Human Resources, Janelle Gale, said the workforce changes were meant to ‘offset the other investments we’re making.’
‘I know this leaves everyone with nearly a month of ambiguity, which is incredibly unsettling,’ she said.
The layoffs were allegedly scheduled for May 20, Meta employees told the Times.
As thousands of workers brace for layoffs, none of them knows whether their employment is about to end abruptly, or if they will be replaced by the very software that tracked them.
‘It’s incredibly demoralising,’ an employee remarked.
If AI ever turns into a ‘worst-nightmare sci-fi scenario’, it won’t be because the technology chose to take over out of the blue. It will be the result of poor human construction, deployment, or governance.
This distinction is significant because it implies that the future is moulded by decisions, laws, and design, rather than being predestined or doomed.
Every real-world AI risk today is human-driven, not machine-driven, and things that feel like sci-fi nightmares are usually failures of power, oversight, or incentives, not rogue intelligence.
Humans design and absolutely control AI; its quality depends on the design, governance, and the people running the systems.
There are three layers: design control – humans determine the architecture, training data, safety rules, and limits. Humans determine where AI is deployed, what it’s authorised to access, and what it can do, and governance control is about laws, audits, and oversight bodies that decide what companies are authorised to build.
AI doesn’t wake up one morning and decide to do something. It runs the instructions and constraints humans give it.
AI cannot act independently in the physical world. It cannot change its own goals. It cannot override safety systems, and it cannot access systems or data unless humans explicitly connect it, and even the most advanced models are pattern-recognition engines, not agents with intent.
However, this is where things get messy, and this is the part people feel worried about, and not without good reason.
Corporations control AI more than the public does, and decisions about training data, safety, and deployment are made by a small number of companies. Governments are behind the curve. Regulation is slow, fragmented and usually written by people who don’t understand the technology, and then bias and power dynamics leak into systems.
If the people building AI don’t understand disability, class, or lived experience, the systems inherit those blind spots.
As for AI becoming aware, that is extremely unlikely because they are made by humans. Deciding to wipe out humanity, possibly because humans are the controllers of the AI. Becoming ‘evil’, maybe not evil, but they could be used to be destructive against humanity.
After all, they are now employed for control and monitoring, and of course, algorithms are always spreading false information, and because the pace is fast, the communication is lacking, and the people building the systems usually act as if they are protected from consequences.
It creates the impression of an unbeatable force, but the reality is more grounded. AI is powerful, but brittle. Impressive, but narrow. Useful, but dependent. It’s not an alien intelligence. It’s a mirror of human systems, including their flaws.
THE MONTHLY MOSQUITO 