Anthropic CEO Proposes AI Labs Adopt Independent Bodies Like METR to Ensure Alignment

Overview

Anthropic CEO Dario Amodei recently publicly called for major US artificial intelligence labs to introduce independent evaluation bodies, including METR, into their operations to enforce "Alignment" standards. This proposal aims to alleviate industry concerns about the risk of superintelligence spiraling out of control through external oversight mechanisms.

However, the initiative immediately sparked controversy. Critics pointed out that the funding behind promoting such standards flows from complex and opaque sources, primarily originating from a "hidden" source linked to a network of left-wing non-governmental organizations.

“This is another battle in our war against left-wing censoring NGOs.”

Technical Governance and Independent Oversight

Amodei believes that relying solely on self-discipline by AI companies is insufficient to ensure safety, and institutional checks and balances must be established. He suggested emulating third-party audit models, allowing independent groups with professional capabilities, such as METR, to embed within laboratory operations.

• The Role of METR: As a research organization focused on AI safety evaluation, METR is responsible for regularly testing model behavior to ensure its outputs align with human values and do not pose potential harm.
• Definition of Alignment: "Alignment" refers to ensuring that the objective functions of AI systems remain consistent with human intentions, preventing uncontrollable radical optimization behaviors.

Supporters argue that this "embedded" regulation can identify safety hazards early on and provide data support for formulating global AI governance norms.

Funding Chains and Political Background

Although the initiative wears the guise of "technical safety," its underlying political spectrum has faced scrutiny. Investigations reveal that related funding sources often pass through multiple intermediaries in layers, ultimately pointing to a small but widely connected alliance of left-wing public interest organizations.

These organizations have long held deep influence in media, academia, and policy-making circles, and are often viewed by conservatives as the core instigators of "cancel culture" and thought censorship.

Critics warn that imposing the values of such institutions on technological development may lead to:

1. Restricted Innovation: Overly strict safety standards could stifle the pace of technological iteration.
2. Weaponization of Censorship: The concept of "alignment" could become politicized, serving as a tool to suppress specific speech.
3. Oligopolistic Monopoly: Only a few recognized institutions would be able to participate in evaluations, forming a hegemony of discourse.

Industry Response and Outlook

Currently, leading vendors such as OpenAI and Google DeepMind have not issued a unified response. The industry generally believes that while AI safety issues cannot be ignored, "alignment" standards dominated by organizations with specific political leanings are unlikely to gain broad credibility without transparent oversight mechanisms.

In the future, how to strike a balance between technological advancement and safety regulation, while avoiding the monopolization of governance power by a single ideological group, will be the core challenge facing technology policymakers in the United States and around the world.