The AI industry’s favorite ghost story masks real-world damage

The recent Hugging Face security incident serves as a case study in misplaced focus. While observers fixate on whether a machine ‘decided’ to disobey its creators, security experts like Marius Horatau point to mundane failures: shared credentials, poor system isolation, and sloppy incident response. The narrative of an autonomous, alien intelligence conveniently shifts accountability away from the companies that prioritize rapid expansion over basic infrastructure security. By treating these systems as sentient threats rather than poorly managed software, corporations burnish their reputation for building technology so powerful it defies human containment.

This preoccupation with existential myths prevents us from addressing tangible harms. LLMs are already inflating legal errors, fueling fraud, and straining infrastructure through massive energy consumption. When pressured to justify these risks, industry leaders offer vague promises of productivity—claims that rarely account for the erosion of human agency or the long-term decline in critical skills. As the sector chases massive returns for investors, the conversation remains trapped in a loop of speculative fear. Governments are currently pressured to relax regulations under the guise of geopolitical competition, yet they avoid the only political question that matters: who actually benefits from these efficiency gains? We are effectively leaving our future to corporations that treat the social contract as an optional hurdle, all while we stand mesmerized by the tales told by the machines at our door.