Daily digest

Today in AI

Wednesday, 19 August 2026

The top 8 of 10 AI stories in plain language, a prompt to try and one AI word to learn — about five minutes.

01
OpenAI Astra: The AI Safety Slowdown
OpenAI is temporarily slowing parts of frontier AI development as increasingly capable agents raise new cybersecurity, monitoring, and control challenges.
Read the story →Source: openai.com
02
AI Agents: The Rise of Claude & GPT
The AI race is shifting from smarter agents to controllable agents with identity, permissions, audit logs, and human oversight.
Read the story →Source: openai.com
03
Claude AI: The Battle for Creator Rights
AI companies need human content to build better models, but creators are asking a bigger question: who owns the right to use it?
Read the story →
04
Microsoft Copilot: The Future of One AI
Microsoft is reshaping Copilot into one unified AI experience while retiring several existing features.
Read the story →
05
Amazon AI: Data Race for Training
A 404 Media investigation tracked rare books to an Amazon facility where books are reportedly scanned for AI-related data and destroyed.
Read the story →Source: 404media.co
06
Replit AI Enterprise Controls: What Changes for Teams
OpenAI is funding 14 independent projects exploring AI policy, jobs, economic opportunity, governance, and societal resilience.
Read the story →Source: openai.com
07
Replit AI: Future of Enterprise Coding
Replit adds admin APIs, audit logs, and workspace controls to help companies manage AI coding at scale.
Read the story →Source: replit.com
08
Claude Mythos: Anthropic’s Cybersecurity AI Expands
Kraken parent Payward has joined Project Glasswing, gaining controlled access to Anthropic’s advanced AI for defensive cybersecurity work.
Read the story →Source: anthropic.com