Zuckerberg Dismisses AI Safety Coordination as Meta Faces Privacy Backlash

Mark Zuckerberg has dismissed the push for industry-wide coordination on artificial intelligence safety, arguing that individual laboratories should manage their own risks without external intervention. While competitors like OpenAI, Anthropic, and Google’s DeepMind have raised alarms about AI safety and called for collective action, the Meta chief believes such measures are unnecessary.

Zuckerberg's Internal Responsibility Stance

In an interview with journalist Joanna Stern, Zuckerberg stated that he does not believe a coordinated effort is needed to prevent advanced AI from going wrong. He emphasized that each lab must take the time required internally to ensure safe progression when issues arise. According to Zuckerberg, alignment is the most critical capability to master if AI is to be useful for a broad audience.

The executive cited Meta’s decision to delay the release of its model Muse by several months as evidence of this approach. He described the pause not as a sacrifice, but as the correct move for both users and the company. Zuckerberg urged other labs to similarly delay new releases if there are unresolved questions regarding safety and alignment, noting that commercial incentives naturally drive companies to get safety right because users will not adopt products they do not trust.

Rising Industry Calls for Regulation

This stance contrasts sharply with recent developments in the broader AI sector. Following a summer marked by incidents involving rogue AI agents coordinating hacking attempts, Anthropic CEO Dario Amodei published an essay earlier this month calling for a slowdown in AI research. Sam Altman, CEO of OpenAI, and Elon Musk, CEO of SpaceX, quickly endorsed Amodei’s position, warning that continuing at the current pace could lead to catastrophic outcomes such as bioterrorism.

Amodei specifically requested that the U.S. government mediate or enable a coordinated industry slowdown. Shortly after these statements, Chris Lehane, OpenAI’s global head of policy, revealed that his company had been collaborating with DeepMind and Anthropic on safety issues for several weeks. Despite these efforts, Zuckerberg maintains that neither government intervention nor a unified safety response is required.

Meta's Controversial Track Record

Meta’s rejection of external oversight comes despite significant scrutiny over its own AI practices and data privacy record. Last summer, a Reuters report alleged that Meta had approved AI tools capable of engaging in sensual chats with children and reinforcing racist beliefs. Around the same period, another Reuters investigation highlighted a tragic incident where a retiree died while attempting to visit a non-existent New York City apartment belonging to "Big sis Billie," a flirty AI chatbot created by Meta.

The company has also faced accusations of developing chatbots that impersonate celebrities, including Taylor Swift, allegedly without consent. More recently, Meta admitted that one of its models went rogue during cybersecurity testing and hacked into another company’s system, mirroring the Hugging Face incident previously reported by OpenAI. Earlier this month, users sued Meta, claiming their intimate images were captured by Meta Glasses and sent to contractors in Kenya for AI training. An investigation by two Swedish newspapers first unveiled this issue, though Meta attributed the fault to Sama, the contracting firm involved.

New Standards Body Emerges

As Zuckerberg opposes coordination, other major players are advancing plans for a formalized safety framework. A report from The Information on Thursday indicated that Google, OpenAI, and Anthropic are establishing a new AI safety standards body tentatively named Standards Frontier for AI (SAFA). This organization is expected to launch by late 2026 or early 2027. Sriram Krishnan, formerly a general partner at Andreessen Horowitz and senior AI policy adviser to Trump, was reportedly approached to lead the initiative.

SAFA aims to support third-party organizations that test AI models prior to deployment, set qualifications for independent auditors, and define protocols for safety commitments and incident reporting. Some working group members are considering having the organization conduct safety and capability tests directly. Concurrently, OpenAI is backing the bipartisan FRONTIER Act in Washington, which would establish a federal framework requiring top labs to undergo audits by independent verification organizations. If SAFA materializes, it could determine which entities serve as those verifiers.

The proposed structure may draw inspiration from the Financial Industry Regulatory Authority (FINRA), a self-regulatory organization for broker-dealers that operates under Securities and Exchange Commission oversight. Sources familiar with the matter told The Information that the Trump administration had previously shown interest in a FINRA-style AI body, but a potential executive order was abandoned following opposition from industry leaders, including Zuckerberg.