AI Radar | July 27, 2026
Today: Cognizant trains 30,000 employees on Claude, Microsoft ships a cheaper security model, and Claude chats are showing up on Google. Read that last one if you use Claude with your business data.
- 01
Cognizant trains 30,000 employees on Claude and builds it into its platforms
The company became a global Anthropic partner and trained more than 30,000 workers to use Claude. Cognizant says one client cut contract review time by 40 percent, and another saved 8 hours a week on risk assessment. If you hire consultants or outsource operations, ask your vendor if they already use AI and how it changes turnaround times.
Read on anthropic.com - 02
Microsoft launches a security AI model that cuts costs in half
The new MAI-Cyber-Flash handles 90 percent of security tasks and leaves only the hardest ones for more expensive processing. Microsoft says that cuts spending by 50 percent.
Read on microsoft.ai - 03
Nvidia invests in Ilya Sutskever's AI startup and multiplies its computing power
Safe Superintelligence, valued at $30 billion, gets a major investment from Nvidia plus access to hardware that grows its capacity 'by an order of magnitude.'
Read on wsj.com Paywalled - 04
Claude user conversations show up in Google search, with no warning
Shared Claude chats turned out to be indexable by Google. That means conversations that looked private can be found on the internet. Before you put sensitive business info into any AI tool, check exactly what data ends up public and how you control privacy.
Read on 404media.co - 05
Kimi releases an open-source AI model with vision and a 1 million token memory
The new K3 is a huge model that understands images and can handle extremely long texts. Its makers say it uses fewer resources than similar models. Because it's open, you can run it on your own servers without depending on a single vendor.
Read on Kimi.ai on X - 06
Librarians run viral workshops on turning off AI in phones and computers
Across the U.S., librarians are holding packed sessions that teach people to turn off Apple Intelligence and Gemini on their devices. They call it digital freedom. If your business uses AI to talk to customers, tell them clearly when they're talking to an AI and how to reach a person.
Read on techcrunch.com - 07
Midjourney improves its image generator with better looks and more personalization
Version 8.2 makes better-looking images and gets better at matching each user's taste as it learns their preferences.
Read on midjourney.com - 08
Grok comes free to Google Sheets, Slides and Docs with built-in AI help
xAI's new add-on lets you answer questions in spreadsheets, write formulas and build presentations right inside Google Workspace. If your team builds reports in Sheets, they can try it to write the formulas.
Read on x.ai