Anthropic Warns Open Model GLM-5.3 Poses Cyberattack Risk as Open-AI Debate Intensifies

MIT Technology Review reported that the dispute is not only ideological. Nvidia CEO Jensen Huang, whose company has a lot to gain if people continue buying their own hardware to run open models, advocates enthusiastically for the United States to develop more of them. Anthropic’s Dario Amodei, who runs a company that could be undermined by cheap, powerful models, advocates for a more cautious approach. David Gray Widder, an assistant professor at the University of Texas at Austin School of Information, said, “I think a lot of the way you come down on the openness debate is influenced by how you come down on the AI debate.”

MIT Technology Review’s The Download newsletter described the warning as the latest flashpoint and said the debate over open AI is more complicated than it seems. The technical difference between open and closed models is straightforward. Closed models, such as those that power ChatGPT and Claude, can be accessed only through a website, app, or API. Open models can be downloaded directly; details about their internal structure, encoded in a long series of numbers known as model weights, are publicly available. Such models are sometimes called open-weight models, to distinguish them from models for which even more information, such as details about the training procedure, is available.

Closed models are currently more capable. As of publication, the top eight models on the Artificial Analysis LLM leaderboard were all closed, though open models were only a handful of months behind in development. But there are several reasons someone might choose an open model. Because open models are downloadable, people can run them on their own computers or hard drives. They can even retrain them to behave differently if they have the expertise and resources. Academic researchers working to understand and improve AI systems almost certainly need open models: with closed models, researchers can do little more than send prompts and observe the responses, while open models allow them to observe the models’ inner workings and the chain of thought they use before answering. Researchers can also combine multiple open models to build new systems, compress them to make them more efficient, and fine-tune them for new applications.

Businesses have reasons to prefer open models as well. Even if they depend on an external provider to run them, open models typically cost less than closed models, according to a recent report from Mozilla. If businesses want to use AI to process sensitive data or proprietary information, running a local model on their own hardware may be the best way to keep their operations secure.

Elizabeth Seger, senior policy advisor at the Tony Blair Institute for Global Change, said the clearest and biggest argument for open models is preventing mass concentration of power and making sure this foundational technology is a public good. “I think we’re at this juncture where the question of who controls the infrastructure that allows AI to operate is as big of a question as who gets to control information on the internet,” she said.

Safety advocates warn that open models can easily be misused by bad actors, while supporters argue that this power should not sit behind the closed doors of just a few big companies and that open models could help spread it more widely. There are a range of ways to handle open models between banning them entirely and releasing them without restriction, but any approach a government or society might take still involves trade-offs between safety and democratization. Deciding whether to support or limit the production of open models requires understanding those risks and benefits in depth.