xAI Wants ‘Lifetime’ Access to Staff Facial Recordings to Train Machine Learning, Report Claims

In a shock move reverberating through the tech world, Elon Musk’s AI startup xAI has reportedly asked employees to give it “never-ending” access to recordings of their faces. The initiative is part of ongoing efforts to improve its AI chatbot, Grok. The revelations—published by Business Insider—suggest that xAI has been using its own staff to help train its systems to recognize and mimic human emotions through facial expressions.

Training Grok with Human Emotion

Central to the report is the claim that, in April, xAI invited staff to participate in a unique data-gathering exercise:

Recording an array of their facial expressions on camera.

According to the report, the footage was intended to enhance Grok’s ability to read emotional cues, now seen as essential to developing AI assistants that can interact in more human-like ways.

While the request was framed as voluntary, there was a major caveat. Employees were asked to sign a consent form granting xAI the right to use their “likeness”—including facial appearance and expressions—for:

• Product development
• Advertising of commercial products
• And indefinitely

What the Consent Form Supposedly Said

Documents viewed by Business Insider allegedly show that xAI presented this consent during layoff negotiations. Key clauses included:

• Immortal consent to use, store, and process employees’ facial image
• Exclusive rights for training commercial AI systems
• Permission for promotional and marketing use
• No expiration or revocation clause, implying permanent consent once signed

Importantly, the document did not provide a way to opt out after signing. It also failed to mention whether participants would be compensated for sharing such biometric data—an omission troubling to both labor advocates and privacy experts.

Facial Data in the A.I. Age: Innovation or Overreach?

Training AI with facial data isn’t entirely new. It’s often based on:

• Public datasets
• Stock images
• Paid actors

However, xAI’s move to harvest this data from employees—especially under ambiguous terms—has sparked intense debate.

“Facial expressions are highly personal, and assigning perpetual rights to one’s likeness raises ethical questions about consent, privacy, and long-term use,”
— Dr. Elaine Morrison, AI ethicist, University of Washington

She adds,

“When an employee hands over rights to their facial data forever, it becomes nearly impossible to control how that data will be used in the future—especially as commercial applications evolve.”

What Is Grok and Why Does It Require Faces?

Grok is xAI’s conversational assistant, integrated into Musk’s social media platform X (formerly Twitter). Positioned as a more free-form, less filtered alternative to ChatGPT, Grok is designed to:

• Handle sensitive topics
• Show emotional intelligence in interactions

To support these capabilities, Grok’s training has gone beyond traditional text-based datasets, venturing into emotionally nuanced facial inputs.

Yet, sourcing that data from employees raises legal, ethical, and workplace challenges.

Legal and Ethical Questions Loom

The idea of staff signing away permanent rights to their likeness flirts with the line between technological innovation and personal invasion.

• In the U.S., biometric laws are fragmented
• States like California and Illinois have stricter rules
• However, there’s no federal law explicitly banning such data collection if consent is granted

But is that consent truly voluntary in a workplace setting?

“This is exactly why consent frameworks in employment should be closely examined,”
— Jennifer Alston, labor rights attorney, San Francisco

“If you’re asking workers to give up control of their face forever, just to keep working, that’s not fair or balanced.”

Further concerns stem from international law. The European Union’s GDPR offers the right to revoke consent and be forgotten. If Grok launches globally, xAI’s practices could come under intense regulatory scrutiny.

The Bigger Picture: AI and the Human Image

xAI’s strategy mirrors a wider trend among AI giants like:

• Meta
• Google
• OpenAI

These companies are exploring emotionally intelligent AI that reads facial cues, tone, and body language. However, few have been as bold—or controversial—as xAI appears to be.

As technology advances, so do the ethical expectations. A growing chorus of industry leaders and policy makers is calling for:

• Stronger regulations
• Clearer data rights
• Stricter guidelines for biometric data collection in AI

Final Thoughts

xAI’s reported request for lifetime access to employee facial recordings—without compensation or clear opt-out terms—raises major concerns in the evolving landscape of AI ethics.

While training Grok to understand human emotion is a forward-looking goal, the path to achieving that must be transparent, fair, and ethical.

In the race to build emotionally intelligent AI, how companies treat the humans behind the code may ultimately define their legacy as much as the AI systems themselves.