AI Round-up: 29 July 2026

OpenAI and Anthropic converged on the need to pace frontier AI development. Both organisations backed the Pacing the Frontier petition, while Reuters reported a wider employee-led call for a US-supported international effort to manage advanced AI risks.

Anthropic also said Claude Mythos Preview had helped researchers identify weaknesses in cryptographic algorithms. OpenAI quietly released the open-source Codex Security CLI for repository scanning and fix verification, and introduced new live and batch transcription models; meanwhile, Reuters tracked another Asian market sell-off as AI investment concerns persisted.