AI Newsroom Scoops Human Reporters at Security Conference

RuntimeWire represents a sharp departure from the wave of low-quality synthetic content that flooded the internet in the early years of the AI boom. Rather than simply replacing human reporting with AI-generated slop, Merket is attempting to break news and compete on speed and sourcing—a model that succeeded visibly when RuntimeWire scooped traditional media at the Black Hat conference. The economics of the operation make this plausible: at $100 per day, the cost structure is so low that even modest audiences can sustain the project.

Merket's approach sits uneasily between journalism and automation. He claims to follow journalistic ethics—contacting sources for comment, linking to sources, issuing corrections—yet he also admits to retracting stories not because they were false but as a personal favor to founders. The legal and factual gatekeeping that RuntimeWire depends on is entirely algorithmic: an AI agent assigns a legal risk score, and Merket publishes stories the AI deems safe without reviewing them first. This arrangement inverts the human-first editorial model that Pete Pachal and Nicholas Diakopoulos consider essential to traditional reporting. Diakopoulos has documented a concerning feedback loop: AI tools (like ChatGPT and Claude) surfaced AI-written sources 16 percent of the time in his tests, which could mean AI newsrooms inadvertently amplify synthetic content by feeding it back to other AI systems.

The broader question—whether this is journalism at all—remains contested. Diakopoulos is skeptical that AI agents can cultivate source trust or handle reporting that relies on human relationships. Pachal sees potential only in narrow domains: aggregation, scraping structured datasets, or live-blogging events like product launches. But Merket's early audience traction (RuntimeWire stories are already circulating on social feeds) suggests the public may not draw the same line between human and synthetic reporting that journalism scholars do.