This Week in AI: Safety Standards, Watermarking, and Agent Memory

OpenAI, Anthropic and Google DeepMind are discussing shared AI safety standards and a possible industry body called SAFA. OpenAI began phased text watermarking under the EU AI Act, while Cohere launched North 2 with cross-session agent memory and a redesigned orchestration harness.

DEV Community

Open source

This Week in AI: Safety Standards, Watermarking, and Agent Memory

OpenAI, Anthropic and Google DeepMind are in talks on AI safety coordination
A proposed industry body SAFA would handle shared testing and audits of frontier models
OpenAI opened opt-in text watermarking for API customers, with EU ChatGPT and Codex next
Cohere North 2 adds cross-session agent memory and admin controls for cost and quotas

#OpenAI#Anthropic#Google#Cohere

Africa demands a say in global AI safety standards

Policy

This Week in AI: Safety Standards, Watermarking, and Agent Memory

Africa demands a say in global AI safety standards

Google, OpenAI and Anthropic Form AI Safety Standards Group

AI agents refactored 300K lines of code in three weeks for $4,000

19 frontier AI models still put Calgary on standard time after Alberta dropped DST