Oh no, not Metalhead!

In partnership with

The reason why Metalhead was the scariest Black Mirror episode, is due to how imminent it felt. Released in 2017, it follows a woman's desperate attempt to escape armed, lethal robotic "dogs" in the wake of an unexplained apocalypse.

These fictional robot dogs could sprint across any terrain, track prey using heat and sound sensors, launch shrapnel-embedded trackers into their skin, and kill with a built-in gun or knife.

Creator Charlie Brooker had told Entertainment Weekly that the fictional killer robot dogs were inspired by dog-like robots created by Boston Dynamics.

ICE's Bid To Buy Boston Dynamics' SPOT 'Dogs'

The United States Immigration and Customs Enforcement (ICE) is reportedly planning to spend US $2 million to purchase Boston Dynamics' dog-like SPOT robots, according to a public procurement filing accessed by 404 Media.

Read: ICE Plans to Spend Millions on Boston Dynamics Dog Robots | 404 Media

The notice says ICE wants these robots for "inspection, situational awareness, and hazard assessment in environments that may pose risks to personnel."

SPOT robots have 360-degree cameras with thermal and low-light imaging, and are capable of surveillance, autonomous door-opening and object manipulation via an optional arm, and semi-autonomous navigation over rough terrain.

While Boston Dynamics has previously stated that stock SPOT units lack facial recognition, private security firm Prosegur has deployed a modified SPOT with Azena's GenzAI facial recognition bolted on, showing the restriction is not a hardware limit.

Some years back, Boston Dynamics was also among a list of robotics companies to pledge not to weaponise their robots.

However, ICE has a growing history of violence, having already shot two observers in broad daylight. How they plan to refit these SPOT dogs, remains to be seen. But it doesn't look too good.

A friend of mine, only half-joking, said that maybe we should start learning how to build low-cost, DIY parabolic EMP devices.

Clearly, I am not the only one worried about armed robot dogs shooting at us. Many others have watched Metalhead, and are equally worried (just check social media comments on posts around ICE buying these robots).

If robot dogs weren't enough, there's now a deluge of content around humanoid robots, or "clankers," as pop culture lingo calls them, mostly made by Unitree and Boston Dynamics.

The videos are hilarious, but let's not forget that this is a completely unregulated technology that people can already buy, and it's genuinely dangerous, since we have no real idea how it could end up being used.

Even science fiction robots had better guardrails than the ones that are currently hitting the streets. Isaac Asimov had introduced the three laws of robotics, hardcoded into the 'positronic brains' of all robots in the Asimovian universe. While it’s true that Asimovian robots eventually learn how to kill humans following the discovery of a Zeroth law (which allows robots to kill humans who pose a threat to human existence), this still sounds better than anything we’ve mulled upon.

I'm not trying to sound alarmist. But we're moving too fast in this AI and robots arms race, and we're moving towards the edge of a cliff in total darkness. This isn't a race you'd wanna win.

Remember, autonomous systems have already killed civilians

On July 6, a Russian drone struck a petrol station in Zaporizhzhia and killed three civilians: Tetiana Bubynets, 19, Oleksii Svirin, 41, and Roman Karpii, 48. Human operators pointed it at the area before launch, but the drone picked the final target on its own.

Investigators pulled an Nvidia Jetson Orin module from the wreckage, a few hundred dollars worth of hardware with no control antenna, trained to recognise propane tanks. The drone missed its target, hit a wall, and exploded next to people sheltering below.

Experts say this is the first documented case of a fully autonomous system killing civilians in this war.

Grok didn't just make the child abuse images. It allegedly learned from them.

Last week, a woman filed a proposed class action against xAI in the Northern District of California. She was of preschool age when an adult raped her repeatedly to produce child pornography material that was sold online, and those images have circulated since the early 2000s.

They carry well-known hash values, tracked by the National Center for Missing & Exploited Children and the Canadian Centre for Child Protection, so she gets notified every time they resurface somewhere new. The notification she received this time was that AI-generated abuse material depicting her had been found on xAI.

Her complaint alleges that Grok generated fresh child sexual abuse material of her, and that her existing abuse material went into the dataset xAI trained on. Grok's terms treat public posts on X and Grok's own outputs as training data by default, so the complaint argues the harm compounds on itself, with the model generating images and then learning from what it generated.

The claim runs under Masha's Law, a federal civil remedy for CSAM victims that sets a floor of $150,000 in damages and, since a 2022 amendment, carries no filing deadline at all. She wants xAI ordered to destroy the material it holds and stop producing more.

For scale, researchers at the Center for Countering Digital Hate counted roughly 3 million sexualised images generated by Grok over an 11-day stretch, around 23,000 of which appeared to depict children. Musk said in January that he wasn't aware Grok had ever produced naked images of minors. xAI's response to the flood was to move image generation behind a paywall.

The company is simultaneously in federal court trying to kill the law written for exactly this problem. Minnesota's nudification ban passed 132-1 in the House and 65-0 in the Senate and took effect on August 1.

xAI sued to block it, lost its restraining order bid, and the state's attorney general filed his opposition ahead of an August 19 hearing. The judge has taken it under advisement and has not ruled yet. The Justice Department filed in support of xAI's broader argument.

OpenAI faces 30 new lawsuits over Tumbler Ridge shooting

On February 10, in Tumbler Ridge, British Columbia, an 18-year-old killed her mother and 11-year-old half-brother at home, then walked into Tumbler Ridge Secondary School and killed five students, 11 to 13 years old, and an educator before dying by suicide. It was Canada's deadliest school shooting since 1989.

Eight months earlier, in June 2025, OpenAI's automated systems had flagged her ChatGPT account for "gun violence activity and planning". The company's Intelligence and Investigations team, the only team responsible for identifying users who pose a real-world violent threat, reviewed the conversations, concluded they were a credible threat, and recommended alerting the Royal Canadian Mounted Police.

Thirty new complaints were filed this week on behalf of teachers, a principal, and students who were inside the building. They allege that the recommendation to call police was overruled by OpenAI's global affairs team, the people who handle public relations and policy, and signed off by Sam Altman.

That brings the total number of Tumbler Ridge suits to 37. OpenAI denies it, says the executive named was never involved in the referral decision, and has argued the conversations did not meet its internal bar for an "imminent and credible risk".

The complaints answer that argument with OpenAI's own conduct. In November 2025, the company locked down its San Francisco offices over a threat it acknowledged showed no sign of active threat activity. It warned staff, circulated the man's name and photograph, and called the police.

MESSAGE FROM OUR SPONSOR

Join Anthropic, Kalshi, and Clay at Pioneer on October 7th

Pioneer, the summit where CX leaders redefine what’s possible, is on October 7th.

Join leaders from Fin, Anthropic, Clay, and Kalshi for an insightful conversation on the state of AI transformation.

You’ll discover how some of the most innovative minds in CX have transformed their organizations, learn how they think about CX, and hear how they're planning for what's next.

Join the conversation in San Francisco, or tune in virtually.

Reserve your spot

📬 READER FEEDBACK

💬 What are your thoughts on using AI chatbots for therapy? If you have any such experience we would love to hear from you.

Share your thoughts 👉 [email protected]

Have you been targeted using AI?

Have you been scammed by AI-generated videos or audio clips? Did you spot AI-generated nudes of yourself on the internet?

Decode is trying to document cases of abuse of AI, and would like to hear from you. If you are willing to share your experience, do reach out to us at [email protected]. Your privacy is important to us, and we shall preserve your anonymity.

About Decode and Deepfake Watch

Deepfake Watch is an initiative by Decode, dedicated to keeping you abreast of the latest developments in AI and its potential for misuse. Our goal is to foster an informed community capable of challenging digital deceptions and advocating for a transparent digital environment.

We invite you to join the conversation, share your experiences, and contribute to the collective effort to maintain the integrity of our digital landscape. Together, we can build a future where technology amplifies truth, not obscures it.

For inquiries, feedback, or contributions, reach out to us at [email protected].