Cybersecurity Startup Raises $47M to Pre-Register Software Packages AI Reliably Invents

SAN FRANCISCO — A cybersecurity startup has raised $47 million in Series A funding after demonstrating that nine leading AI coding tools — including Cursor, GitHub Copilot, and Gemini CLI — reliably recommend installation of the same nonexistent software packages with up to 100% reliability, a consistency the company's founders describe as "the most predictable real estate market we have ever seen."

The firm has pre-registered approximately 4.7 million package names on npm, PyPI, and GitHub corresponding to repositories AI coding assistants routinely hallucinate. Rather than leave the names available to attackers — who researchers this week showed could serve malware by registering them first — the company fills each package with audited, empty code, intercepting what academic researchers called a "pull-based prompt injection at scale."

The business model requires AI coding assistants to continue recommending the same nonexistent packages indefinitely. Asked at a closed-door investor demo whether improvements in model accuracy posed a risk to the company's core offering, the chief executive said the concern had been "marked amber" in the risk register.

"We looked at the nine-year hallucination trend line," he said. "We moved it back to green."

The company's patent-pending framework also pre-registers packages AI systems have not yet recommended but are statistically likely to invent, allowing the firm to secure the "full hallucination surface" of the three largest coding assistants before any model output references them.

The firm's Series A deck describes its market thesis in two sentences: "We are not in the software business. We are in the wrong-answer business."

"If the models ever get this right, we have priced that in at approximately zero," a spokesperson said.