OpenAI published an in-depth update on AI progress and global safety preparedness, forecasting that current systems are already “80% of the way” to the capability of an AI researcher. The company called for coordinated international frameworks as intelligence scales toward superintelligent territory. Official post: AI Progress and Recommendations.

• Performance milestones: OpenAI said its latest models already outperform top humans on complex intellectual benchmarks and could start achieving scientific discoveries by 2026, with larger breakthroughs by 2028.
• Exponential efficiency: Intelligence costs are reportedly falling 40× per year, accelerating capability growth.
• Global cooperation: OAI emphasized the need for partnerships with governments, labs, and safety agencies to prevent misuse—such as bioterrorism or uncontrolled self-improvement.
• Resilience & standards: The post calls for international safety baselines, an ecosystem akin to cybersecurity, and continuous monitoring of AI’s real-world impact.

Why it matters: OpenAI’s latest communication reframes the next few years as a pivotal period before superintelligence emerges. The company is signaling that collective global safeguards may be the only path to manage the risks of self-improving AI—making this one of the strongest public calls yet for preemptive coordination.