Google DeepMind's Internal Audit Finds No Measurable Acceleration in AI Research Progress

Google DeepMind published its first comprehensive internal assessment of AI-driven machine learning research on October 14, 2026, finding no measurable acceleration in AI progress that can be credited to AI itself. Crypto Briefing reported the audit appears inside the Gemini 4 Argon Safety & Security Report.

Inside the Gemini 4 Argon Audit

Researcher Haydn Belfield led the work, which tracked frontier Gemini capabilities with a metric called the General Capability Index, or GCI. Across 123 models and 92 benchmarks, that score has climbed by approximately 0.1 GCI points each month. The report says thorough statistical analysis shows no significant structural speedup after April 2026.

For public models, the monthly change before April 2026 was 0.091 points. After that date, it was 0.118 points. The report found no substantial difference between the two. Gemini 4 Argon itself landed marginally above the expected trend line, a bump the report attributes mainly to more compute and organizational investment rather than the model accelerating its own development.

AI tools contribute to writing over 75% of production code at DeepMind, and 88% of code changes still need human review. Fully delegated changes, where AI handles the work end to end, rose from 7.5% to 11.9%. Surveys of 12 senior DeepMind researchers help explain the pattern: automation works best on clearly defined tasks where the answer can be verified, and struggles with open-ended scientific ideation and complex planning.

DeepMind also ran its agents through RE-Bench, an internal benchmark of 74 realistic machine learning engineering tasks. The Argon agent solved roughly half of them, while Gemini 3.7 Flash managed 27%. A separate review of 1,389 Argon agent runs found that success rates fell notably on complex tasks when no human was involved.

Google DeepMind's Research Partnerships

According to akta.pro company records, Google DeepMind is Alphabet's AI research division, founded in 2010 and headquartered in London, building frontier multimodal foundation models, generative media and scientific AI. The same records list a multi-year AI research partnership with film studio A24 announced in June 2026, backed by an approximately $75 million Alphabet equity stake, and a multi-year strategic partnership with HSBC announced at Google Cloud Summit London 2026 targeting over 200 new AI use cases over two years.

Earlier Google DeepMind Announcements

On October 14, 2026, Singapore's GovTech and Google DeepMind announced a collaboration to build on-device AI models based on Gemma to detect conversational scams in private messages without sending data to the cloud, Crypto Briefing reported. No launch date or performance figures were disclosed.

On October 12, 2026, Google DeepMind introduced SynthID Bio, a hidden watermark for AI-designed proteins that can be verified later, according to The Times of India. The watermark does not impair protein function and aims to track design provenance.

Also on October 12, 2026, Google DeepMind and ICRISAT announced a partnership to use AI to accelerate research on jowar, a crop supporting over 500 million people, NDTV Profit reported. DeepMind's AI Readiness Chief Lila Ibrahim said AI can help speed up jowar research.

What to Watch in AI Research

The share of fully delegated code changes is one trend to watch, since its move from 7.5% to 11.9% is the closest thing to an automation trend line in the report. A sustained break above roughly 0.1 GCI points per month, without a matching jump in compute, would be the signal the self-improvement camp has been waiting for.