AI Daily
Daily AI news updates, published each evening.
22 April 2026: Meta Records Keystrokes, Mythos Under Fire and SpaceX Bets $60B on Cursor (AM)
Meta begins recording employee keystrokes for AI training as Anthropic's Mythos is breached and SpaceX discloses a $60B option to…
AI Daily: Moonshot AI launches Kimi K2.6 with 300-agent swarms, ChatGPT ships Images 2.0
Today's AI news: Kimi K2.6 launches with 300-sub-agent orchestration, ChatGPT ships Images 2.0, and NeoCognition raises $40M to build humanlike…
21 April 2026: The AI That Can Hack Your Browser, Ads Inside ChatGPT, and Half a Billion for UK AI
Anthropic's Claude Mythos Preview finds zero-day vulnerabilities autonomously. OpenAI launches CPC ads in ChatGPT. The UK opens a £500m sovereign…
21 April 2026: OpenAI Launches GPT-Image-2, Adobe Goes Agentic, FCA Names Eight Banks for AI Testing (AM)
OpenAI launches GPT-Image-2 with near-perfect text rendering, Adobe unveils agentic CX platform, and the FCA confirms eight banks in AI…
NSA Deploys Anthropic’s Mythos AI vs Pentagon — 20 Apr 2026
The NSA is running Anthropic’s Claude Mythos on classified networks, agentic shoppers outspend humans at US retailers, and Coinbase builds…
AI Daily 19 Apr: Claude Design Lands, Agents Can Spend
Anthropic launches Claude Design and Opus 4.7. AI agents gain spending power via Mastercard. World ID 4.0 fights deepfakes. FCA…
AI Daily 18 Apr: Claude Opus 4.7 and Anthropic’s Compute Bet
Claude Opus 4.7 is live with sharper vision and better code. Anthropic also launches Claude Design and backs a multi-gigawatt…
AI Daily 17 Apr: Mythos Shakes DC, Opus 4.7 Goes Live
Anthropic's Mythos enters Washington, Opus 4.7 goes live, Google upgrades Deep Think, and Meta launches its first closed-source model today.
AI Daily 16 Apr: GPT-5.4 for Cyber Defence, Canva, DeepL
OpenAI launches GPT-5.4-Cyber with $10M in API grants for security firms. Canva ships full AI design generation. DeepL announces real-time…
AI Daily 15 Apr: Claude Code Rebuilt, UK Flags AI Threats
Anthropic rebuilds Claude Code for parallel agents, OpenAI launches a restricted cybersecurity model, and the UK's AISI publishes alarming Mythos…