Surfshark Engineer Calls AI Extinction Warnings a Marketing Tactic, Cites Deepfake Fraud and Privacy Risks

The warnings had returned over the past week, as technology giants called for pacing the frontier of AI development and researchers expressed fears of AI-driven human extinction. Kaciulis told TechRadar that the latest debate around AI threatening humanity is clearly a marketing move. He said AI companies use the same rogue-AI rhetoric every few months, and that it is almost identical each time.

Kaciulis argued that focusing on a science-fiction future blinds consumers to problems generative AI is already creating. He described the extinction threat as fictional, closer to a Skynet-style scenario than the automated scams and intimidation that are already happening. In his view, AI models are not necessarily becoming independently dangerous, but they are giving bad actors easier access to sophisticated attack methods.

The financial cost is already measurable. According to Surfshark data cited by TechRadar, deepfake fraud has caused $2.19 billion in losses globally, including $149 million in the UK alone.

Kaciulis also pointed to an environmental cost that often sits outside apocalypse narratives. He said one of the most immediate risks from AI is the wildlife and land lost to data centers, the large amounts of energy they use and the water needed to cool them. Surfshark's own research estimates that a single ChatGPT query uses around 2Wh of energy on average, enough to run a 40W desk fan for three minutes. Across hundreds of millions of daily queries, the impact adds up quickly.

Beyond fraud and energy consumption, Kaciulis suggested the AI industry might be projecting future dangers to obscure a less dramatic possibility: large language models may be plateauing. He said the bigger danger for AI companies may be that no valuable scaling is possible anymore with the current state of LLMs, and that they may be as efficient as they ever get. He added that it remains unclear whether newer models are actually better at performing requested tasks or simply better at imitating the responses users expect.

That uncertainty creates a privacy risk, according to Kaciulis. He warned that people do not really understand how chatbots work, or that what they pass to one may be accessible to the company behind it. Because chatbots ask confident follow-up questions while helping with research or tasks, users can forget they are interacting with a corporate data-gathering machine.

For now, Kaciulis's advice is not to prepare for an AI apocalypse but to protect money and personal data from the threats that are already present.