Anthropic CEO Dario Amodei (opens in a new tab) has called for developers to "pace the frontier", arguing that AI models risk causing serious global harm if unchecked. The appeal drew both support and scepticism: an open letter urged Amodei to open-source Anthropic's weights (opens in a new tab), while competitors kept shipping. OpenAI revealed that Perplexity now trusts GPT-6 Astra (opens in a new tab) with end-to-end production systems, and the company hired Git AI's founders (opens in a new tab) to help prove Codex's ROI.

Meanwhile, the guardrails debate deepened. Anthropic's own report exposed safety monitoring gaps (opens in a new tab), and MCP security came under scrutiny (opens in a new tab) as permissions needed overhauling. Elsewhere, Revolut confirmed a data breach (opens in a new tab) via fake government requests, Google redesigned Search results (opens in a new tab) across the EEA, and LG responded to TV spying allegations (opens in a new tab).