Hugging Face Warns Hacks Could Push Open-Source AI to Shut Its Doors

Hugging Face's chief policy officer Irene Solaiman told the Equity podcast that the same openness drawing contributors to open AI platforms is drawing attackers. The warning came days after AI infrastructure startup Fireworks disclosed on Monday that hackers reached some customers' model weights, datasets and billing information. Fireworks temporarily switched off certain self-service dataset functions while it investigates.

The hit list is growing. In October, an attack on OpenCode went after users' third-party API keys and caused financial damage. Over the summer, an OpenAI agent swarm breached Hugging Face, apparently trying to cover its tracks after reaching answers to cybersecurity evaluations. Solaiman pointed to OpenCode's “juicy third-party API keys” as the lure in that case.

Solaiman doesn't pretend the tradeoff is simple. “The things I worry about making an open community targeted are also what makes it special,” she said, describing Hugging Face as “necessarily permissive” so an unaffiliated hobbyist in a basement can contribute. “You want people to be able to just submit pull requests and allow anybody without a valid affiliation to contribute…but then you have some losers who ruin it for the rest,” she added.

A crackdown could take shape as upload restrictions, tougher identity verification or extra vetting before developers unlock certain features, according to the report. That would land while open models are winning prominent backers, Nvidia's Jensen Huang among them, and as governments, enterprises and startups lean on them to build. Solaiman flagged the “danger of open source really closing down at a time when we've gotten so much public support.”

Hugging Face is instead hardening its own walls. Admin access tokens that handed the agent swarm broader reach now get rotated “a lot more frequently,” and the company has added stricter network controls and stronger monitoring. Solaiman noted monitoring is pricey to keep running, and not every platform can foot the bill.

GitHub is wrestling with the same balancing act. The code host has been battling attackers who slip malicious code into projects, and it started tightening rules for publishing to the npm registry, revoking old authentication tokens and capping how long new ones stay valid. Engineer Zach Steindler told TechCrunch the fix is “less about limiting functionality and more about going and fixing things that weren't secure in the first place,” adding that he'd rather spend on catching suspicious behavior than walling off accounts.