Question
By 2028-06-30, will a top-five AI developer by revenue other than Anthropic publish an AI R&D automation index that rates its own AI R&D work on Epoch AI's Automation Level (AL0-AL5) scale?
Current Status and Strict Resolution Criteria
As of mid-November 2026, no developer other than Anthropic publishes ratings of its own AI R&D work on Epoch AI’s Automation Level (AL0–AL5) scale. The resolution criteria for this question are highly specific: an eligible top-five developer (OpenAI, Google DeepMind, xAI, or Meta) must explicitly adopt the AL0–AL5 scale and apply it to its own work. Related disclosures, such as third-party audits by METR How an assessor could investigate evidence of ... https://metr.org/blog/2026-10-14-ai-rd-acceleration-assessment-proposal/, general capability evaluations, or proprietary internal metrics, do not qualify.
OpenAI's Trajectory
OpenAI is the most plausible candidate to adopt the AL scale. The company has a stated goal of developing an "automated AI researcher by March of 2028" and explicitly notes that it will evolve its transparency approach as measurement techniques improve Research acceleration: The view inside OpenAI Research acceleration: The view inside OpenAI. However, OpenAI already uses a different Epoch framework—the six-phase R&D taxonomy—to classify its agent tokens Research acceleration: The view inside OpenAI. Because both the taxonomy and the AL scale originate from the same Epoch proposal Toward an O*NET for AI R&D - Epoch AI - Substack, OpenAI’s decision to use the taxonomy but omit the autonomy levels suggests a revealed preference against the AL scale. While adding AL ratings to a future transparency report would be a minor technical lift, the company's current trajectory points away from it.
Google DeepMind and Other Competitors
Google DeepMind tracks related automation measures, such as the share of AI-written code and fully delegated pull requests Gemini 4 Argon: The Model and Safety Cards - Zvi Mowshowitz, and CEO Sundar Pichai has openly discussed how AI is accelerating Google's model development. However, Google relies heavily on its own internal vocabulary and frameworks, as seen in its recent Argon and Hydrogen model cards Gemini 4 Argon assessment finds no significant AI R&D ... Gemini 4 Hydrogen - Model Card. It actively uses proprietary ML R&D safety and automation frameworks, reducing the likelihood that it will pivot to a scale popularized by Anthropic. Meanwhile, Meta and xAI have shown no indications of moving toward this specific kind of disclosure Advanced AI Scaling Framework - Meta AI xAI Frontier Artificial Intelligence Framework (June 30, 2026).
Standardization Pressures vs. Proprietary Preferences
With about 19.5 months remaining until the June 2028 deadline, there is ample time for industry norms to shift. Anthropic has publicly invited other frontier developers to publish comparable measures Measurements for understanding the pace of AI ..., and organizations like METR or future regulatory mandates could push labs toward a shared standard. Nonetheless, emerging policy proposals tend to focus on confidential regulatory filings or third-party audits How an assessor could investigate evidence of ... rather than public disclosures on the exact AL scale. Weighing this against related measurement questions, I put particular weight on other leading labs’ continued heavy reliance on their own proprietary safety frameworks . Ultimately, the strict requirement for self-published ratings on this exact scale, combined with the strong preference among top labs for proprietary metrics, leads me to a 26% probability.
Weighing this against related measurement questions, this was nudged down to 26% as other leading labs continue to rely heavily on their own proprietary safety frameworks .
Ask a followup
Sign in to run · $20 free credit, no card · every claim cited