Tech Giants Will Form Industry-Led AI Safety Body
Google, OpenAI, and Anthropic plan to launch an independent standards group for model evaluation by late 2026.
Updated on Sept. 25, 2026 in Artificial Intelligence
Live Poll
Do you trust private industry groups to set safety standards for artificial intelligence?
Google, OpenAI, and Anthropic are planning to launch the Standards Authority for Frontier AI (SAFA) by late 2026 or early 2027. The proposed industry-led organization aims to establish testable benchmarks for AI safety that developers can apply to models before commercial release.
Why it matters
The group intends to address perceived gaps in technical depth and funding within existing government evaluation bodies. By creating independent standards, the firms seek to formalize safety protocols for the most powerful AI systems currently in development.
The proposed body plans to move beyond high-level safety pledges by creating quantifiable, testable benchmarks for evaluating frontier AI models. These metrics are intended to provide a technical framework for assessing model performance before products are released to the public.
The players
Sriram Krishnan
A former White House AI policy adviser and current general partner at Andreessen Horowitz.
Demis Hassabis
The co-founder and CEO of Google DeepMind who proposed the creation of the safety body.
A major technology company developing large-scale artificial intelligence models and infrastructure.
OpenAI
A research and deployment organization focused on the development of frontier artificial intelligence.
Anthropic
An AI safety and research company known for building large language models with a focus on steering and control.
The details
SAFA is designed to act as an industry-led entity rather than a formal government regulator. It would provide the infrastructure necessary for developers to conduct rigorous safety evaluations, an area where planners argue current public sector efforts face challenges regarding technical depth and resource allocation. Sriram Krishnan, a former White House AI policy adviser, is being considered to lead the organization.
Timeline
January 2025 to June 2026: Sriram Krishnan served as a White House AI policy adviser.
July 2026: Demis Hassabis proposed the creation of a US-based AI standards body.
Late 2026 or early 2027: Google, OpenAI, and Anthropic aim to launch SAFA.
The Tech Race
This initiative follows the proposal for a national standards body in July 2026 and marks a shift toward industry-coordinated evaluation. It sits in direct contrast to the U.S. AI Safety Institute, aiming to fill perceived technical gaps in current public-sector testing capabilities.
The impact on the public depends on the adoption of these standards by developers before commercial AI tools reach the market. The timeline for when these benchmarks will influence product release cycles remains tied to the organization's launch in late 2026 or early 2027.
The takeaway
This effort underscores the growing pressure to standardize safety testing for frontier models outside of government channels. Readers should track the formal appointment of leadership and the publication of the first technical testing benchmarks expected in 2027.
Further reading
For more on how industry leaders are approaching regulation, see the latest developments in Artificial Intelligence.
Source note: This article includes information reported by News18.
Live Poll
Do you trust private industry groups to set safety standards for artificial intelligence?
