Cybersecurity in the AI era is no longer a separate beat. Models can write convincing phishing emails, clone voices, scale social engineering and find vulnerabilities in code at a speed defenders have to match. This section covers both sides of that arms race. We track AI-powered attacks already showing up in the wild, profile the security startups using LLMs to harden detection and response, and translate dense research from labs like Anthropic, Google DeepMind and the major US universities into briefings security leaders can actually use. You will also find practical guides on protecting prompts and training data, evaluating vendor risk, and building an AI usage policy that holds up the first time something goes wrong. Pragmatic, sourced and never alarmist.

Read more from the ChatGPT and Hardware sections, or browse the free developer tools built by the newsroom.