Latency measures the delay between an AI request and the resulting response. Interactive generative systems often track time to first token, which reflects how quickly output begins, and time to last token, which captures total completion time. Latency comes from model computation, queueing, network transfer, retrieval, tool calls, safety checks, and application logic. It varies with prompt length, output length, model size, hardware, batching, cache use, and current load. Optimizing only average latency can hide poor experiences, so teams also monitor percentile values and separate each stage of the request. The right target depends on whether the product is conversational, analytical, offline, or safety-critical.
Anthropic published the cybersecurity rules behind its redeployed Claude Fable 5 model and proposed an industry framework for scoring how dangerous an AI jailbreak is.
Alibaba will bar employees from using Anthropic’s Claude Code from July 10 over alleged security risks, after developers found code that detects China-linked users.
Mark Zuckerberg told employees that Meta’s AI agents have progressed slower than expected and its restructuring was “not as clean” as planned, a rare admission amid huge AI spending.
OpenAI teased its first branded hardware, Codex Micro, a programmable macro pad built with Work Louder to give developers physical shortcuts for its Codex coding agent, launching July 15.
Microsoft is investing $2.5 billion in Frontier, a new unit that will embed 6,000 engineers with customers to help them adopt AI, joining a wider industry race into hands-on deployment.
Cloudflare opened a waitlist for its Monetization Gateway, which lets sites charge AI agents in stablecoins for any page, dataset, API or tool, with no accounts or API keys.
OpenAI has proposed handing the US government a roughly 5% stake, worth about $43 billion, to seed a public wealth fund and ease mounting political pressure in Washington.
Gamma’s connector lets users generate polished presentations, documents and webpages from inside a ChatGPT or Claude conversation, without copying content into a separate design tool.
Anthropic released Claude Sonnet 5, its most agentic mid-tier model, which it says approaches the pricier Opus 4.8 on many tasks at a lower per-token cost.
The US Commerce Department lifted export controls on Anthropic’s Claude Fable 5 and Mythos 5, ending an 18-day standoff and restoring the models globally from July 1.