Is open-source AI safer or riskier than closed/proprietary AI?

Open-Source vs. Closed AI: There's No Simple Answer — It Depends on What Kind of Risk You're Worried About

This is genuinely contested, and the honest answer is "it depends on which risks you weight most heavily." Here's how the debate breaks down:

The Case That Open-Source AI Is Riskier

Irreversibility of release. Once weights are public, safety measures can't be undone. As Anthropic puts it, drawing on UK AI Security Institute analysis: the same openness underpinning open models' benefits precludes many of the safety measures that closed model developers can use to detect and disrupt misuse, iterate on safeguards as vulnerabilities emerge, control user access and withdraw models. Once open-weight models are released, these options are lost permanently: safeguards can be removed, and copies can be downloaded, redistributed, and run on private systems beyond monitoring.

Guardrail removal is getting easier. A concerning trend is "abliteration" — a technique for stripping safety training out of open models. By tweaking model weights, people can take away the model's ability to say "no." Hugging Face, which hosts open-source AI models, currently lists over 6,000 abliterated models, compared to about 600 in 2024. This is happening as capabilities rise: in 2026, open-weight AI models possess advanced capabilities not far behind their proprietary counterparts, and getting rid of open-weight models' guardrails, which used to take time and deep expertise, has become dramatically more accessible and popular.

National security / geopolitical concerns. The NTIA's official report notes: strategic countries of concern would not need to make the same investments in semiconductors and other hardware in order to produce similarly high performing AI models with the wide availability of model weights. More broadly, releasing weights could also engender harms and risks to national security, equity, safety, privacy, or civil rights through affirmative misuse, failures of effective oversight, or lack of clear accountability mechanisms.

Empirical safety comparisons. Some studies do find open models lag on safety: several studies demonstrate that closed models typically demonstrate fewer safety and security risks, compared to open source. Anecdotally, DeepSeek-R1 has a 100% jailbreak success rate on the HarmBench dataset.

The Case That Open-Source AI Is Safer (or at Least Not Clearly Riskier)

Closed models aren't actually secure by default. The "closed = safe" assumption has real cracks: closed models still demonstrate weaknesses and are also vulnerable to jailbreaking techniques, with safeguards easily bypassed through simple manipulations like fine-tuning via accessible services, prompting the model to repeat a word, applying a cypher, or instructing the model in another language — and completely preventing models from exhibiting undesirable behaviors might not even be possible. Researchers conclude: it is not clear that closed models are definitively "safer" than open-source models, and gaps will likely narrow over time as jailbreaking attacks strengthen and open safeguarding methods improve.

"Closed" doesn't mean "unleakable." Closed models can be compromised via hacking, leaks, reverse engineering, or duplication — for example, Mistral's 70B parameter model was leaked by one of their early customers.

Transparency enables scrutiny. Openness allows outside auditing that closed systems don't permit: open source lowers barriers to entry for new developers and researchers, and encourages a culture of sharing, collaboration and learning, helping to build a more inclusive and knowledgeable AI community, while also allowing for outside scrutiny and auditing. On bias specifically, civil rights groups value this: commenters have highlighted the importance of transparency and oversight enabled by open model weights in attempting to fight bias and algorithmic discrimination, and there's a long history in the civil rights community of collaborative testing that wide availability of model weights enables to continue.

The "vaccine" argument. One provocative framing: exposure to open-source risk forces society to build defenses. Open-source AI can be seen as a "vaccine" for human society — since it may be used maliciously, society is forced to strengthen its defenses, improving overall security. Otherwise, if only closed-source, aligned AI exists, people might become relaxed, and if that closed-source AI fails alignment, is maliciously altered, or stolen by bad actors, humanity could face catastrophic consequences.

Concentration risk. A notable 2026 development: over 270 companies (including Microsoft, Google, Amazon, Meta, OpenAI, and NVIDIA) published an open letter arguing that open-weight AI models are essential to American AI leadership and that concentration among a small number of closed-model vendors constitutes a systemic risk in its own right. Even Anthropic's CEO, while more cautious about frontier open-weight releases, acknowledges it is irrelevant whether models are released with open weights — the most dangerous model may be one trained in secret and handed only to a military or intelligence agency for use in drones or surveillance and repression.

The Nuanced Middle Ground

Most careful analyses converge on: it's not "open vs. closed" that determines risk — it's capability level, deployment context, and specific use case. The NTIA report itself concludes with caution rather than a verdict: this Report provides relatively little insight to future releases of dual-use foundation models with widely available model weights; without changes in research and monitoring capabilities, evidence of risks that would justify restricting model weights might only arise after AI models with those capabilities — closed or open — have already been developed or released.

The emerging consensus among many researchers is that most risks apply to both paradigms, though they manifest differently: there are real risks associated with open-source models, though these are sometimes exaggerated — and most concerns about open sourcing near to mid-term AI models are also pertinent to closed source models.

Bottom Line

There's no consensus "open is safer" or "closed is safer" — it's a genuine trade-off between different risk profiles, and current policy debates (including a major 2026 dispute between AI companies over exactly this question) reflect that this is still very much unresolved.