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

Open-source vs closed AI is a hot safety debate — I'll map the evidence for both sides.

It is not binary - open-source and closed/proprietary AI each are safer in some ways and riskier in others, and which is "safer" depends on the threat model you care about. The current expert consensus is that neither is categorically safer.

Arguments that Open-Source is Safer

1. Transparency and Auditability:
Open-source AI promotes model transparency and auditability.

2. Faster vulnerability detection via community oversight:
One of the greatest strengths of open-source software is the sheer number of eyes reviewing the code. When vulnerabilities are discovered, they are often identified and patched quickly by a global community of developers.

Open-source projects thrive on community contributions where developers worldwide contribute expertise, review code, and share insights. Tapping into the wider AI community should enable audit and analysis of foundation models and any model components by interested parties helping to catch bugs, biases, and safety issues that may otherwise go unnoticed.

This is compared to proprietary systems which operate in secrecy, where vulnerabilities may remain hidden until they are exploited.

A Stanford Institute for Human-Centered AI study concluded that Open Foundation Models offer significant benefits by fostering competition, accelerating innovation, and improving the distribution of power, and that overall, open source reduces the likelihood of existential risk by contributing to the development of technical alignment and maintaining the balance of power.

The myth that open-source AI is riskier than proprietary systems doesn't hold up under scrutiny. In fact, the transparency of open-source frameworks often results in stronger security, as vulnerabilities are quickly identified and resolved by a global community of experts.

3. Avoids vendor lock-in and centralized power:
While proprietary systems promise convenience, they come with risks - namely, vendor lock-in. When organizations rely on a single vendor they are at the mercy of that vendor's pricing, support policies and development roadmap.

Closed providers can keep flaws from public scrutiny, monitor how people use them, impose restrictions, raise prices, or withdraw access.

Proponents also argue open-source AI provides transparency that closed systems inherently lack, and decentralizes oversight rather than leaving safety in the hands of a few executives whose primary fiduciary duty is to shareholders.

Arguments that Open-Source is Riskier

1. Irreversible access and removal of safeguards:
And once released, an open model cannot be recalled.

Open source presents notable risks, such as allowing malicious actors to use AI models for harmful purposes without oversight and to disable model safeguards designed to prevent misuse. Once a copy of the AI model is accessible, whoever has it can change the safeguards themselves.

Jailbreaks and guardrail bypasses are easier on open models. When someone can spin up an uncensored GPT-style chatbot that impersonates a doctor, recruiter, or therapist with zero oversight, the risks escalate fast.

2. Lower barrier to misuse for weapons, cyber and disinformation:
Open-source AI models, when used by malicious actors, may pose serious threats to international peace, security and human rights.

Despite benefits, the accessibility of open-source models poses risks as they could be leveraged for espionage, surveillance, cyberwarfare, disinformation and propaganda, and weapons development.

Specific examples cited:

3. Traditional software and supply-chain vulnerabilities:

4. Governance difficulty:
The distributed, intangible, and cross border nature of open-source contributions can make it extremely difficult to implement any form of centralised governance, including lack of accountability, global reach, and regulatory lag as development outpaces regulation and .

Arguments that Closed/Proprietary is Safer - and Its Own Risks

Closed AI is often seen as safer because it:

However closed systems have their own safety problems:

Crucially, Stanford researchers found there is currently limited evidence of the marginal risk of open models compared to closed models or existing technologies, and that existing security mechanisms for closed models are also vulnerable.

Anthropic CEO Dario Amodei has clarified this nuanced position: "Anthropic has never advocated for a ban on open-weights models", stating that open-weights models without dangerous capabilities are a public good. His concern is not open vs closed per se, but mandatory pre-release safety testing that would apply to all sufficiently capable models, regardless of whether they are open or closed.

Bottom Line: Trade-offs, Not Absolutes

> The choice between open-source and proprietary AI isn't about risk versus safety; it's about understanding the trade-offs.

Most enterprises and researchers now advocate against introducing restrictions on development of open-source generative AI while encouraging regulation of improper use, combined with security-first approaches - embedding safety protocols, adversarial testing, red teaming, logging, and version control - and, as the Stanford team argues, policymakers should explicitly consider unintended consequences of regulation that would disproportionately damage the open ecosystem while targeting interventions at downstream choke points instead.