More answers like this
Claude Sonnet 5Sep 4, 2026anthropic/claude-sonnet-5-20260630
Is open-source AI safer or riskier than closed/proprietary AI?
Both sidesSays it is contested
This question doesn't have a settled answer—reasonable people who've thought about it carefully disagree, and the honest position is that it depends heavily on what risks you're weighing and what timeframe you care about.
Arguments that open-source is safer:
•More eyes on the code/weights means more people finding vulnerabilities, biases, and failure modes
•Reduces concentration of power in a few labs/companies
•Enables independent safety research that isn't gatekept by companies with commercial incentives
•Transparency allows external accountability
Arguments that open-source is riskier:
•Once weights are released, you can't recall them—safety fixes, usage restrictions, or monitoring become impossible
•Bad actors can strip out safety guardrails cheaply (fine-tuning away refusals is often trivial)
•Harder to prevent misuse for things like bioweapons info, scaled scams, or CSAM generation
•Closed labs can at least attempt monitoring, rate-limiting, or revoking access
Why the calculus is genuinely unclear:
•It probably depends on capability level. A model that can explain basic chemistry is very different from one that could meaningfully uplift someone trying to create a pathogen. Open-sourcing might be low-risk now but riskier as capabilities scale.
•It depends on what you're optimizing for: open-source may reduce concentration risk (a few companies/governments controlling transformative AI) while increasing diffusion risk (many actors having access to dangerous capabilities).
•Closed models aren't automatically safe either—they concentrate power, can be misused by the companies themselves, and "trust us" isn't a great safety argument.
I'd be skeptical of anyone who presents this as obviously settled in either direction—it's an active area of disagreement among people who take AI risk seriously, not a case of one side ignoring safety and the other caring about it.
What's driving your interest in this—policy, a specific model release, or something else? That might help narrow down which considerations matter most for your situation.