Open or closed? The fight over how AI models should be released
Should powerful AI models be open for anyone to download and inspect, or kept behind locked APIs? Meta and others champion open models; OpenAI and Anthropic keep their best ones closed, citing safety.
The summary above is a neutral framing. Below, each side reports the same story in its own words — judge for yourself.
Backers argue open models bring transparency, far cheaper inference, faster innovation and no vendor lock-in — and that openness, not secrecy, builds trust and broad access.
The closed camp argues frontier models are too risky to release freely — easier to misuse — and that controlled APIs offer better security, accountability and a business model to fund safety research.
All rights reserved by the original publishers.
Excerpts are shown for comparison and commentary. Visit each source above to
read the full report.