The week in AI: week 43, 2026
October 19 – 25, 2026 · This week so far
Safety
ConnectWise breach exposed thousands of companies to spying
Microsoft says China-linked hackers stole ConnectWise credentials and used AI agents to target about 2,250 organizations. ConnectWise paused its cloud backup service.
Safety
OpenAI and Google AI models spot critical PHP bug affecting many sites
GPT‑5.6 Sol and Gemini 4 Argon identified a severe PHP flaw; a GitHub tool now lets operators test their servers.
Business
Anthropic plans $750 billion compute spend
Anthropic’s prospectus shows a $750 billion investment in AI hardware as it prepares for a possible $2 trillion IPO.
Safety
Army Cyber Command tests AI agents that spread like worms
ARCYBER’s ECHOTRIBBLE program aims to let AI agents move across networks and disable targets with minimal human control.
Business
Javo sales AI lifts online revenue by 10% in field trial
A sales chatbot trained on 4.6 million commercial chats raised revenue per visitor 10% and matched human agents on conversion.
Safety
US bans DeepSeek AI tools in federal civilian agencies
The Department of Homeland Security ordered all civilian agencies to remove DeepSeek products on Oct. 19, 2026, citing security threats.
Models
MiniMax releases M3.1 Flash model at about $0.22 per task
The Chinese firm MiniMax launched M3.1 Flash, a cheaper but stronger reasoning model, with open weights slated for Oct 23, 2026.
Tools & agents
Major US banks block AI agents that use customer passwords
Morgan Stanley, JPMorgan, Bank of America, Wells Fargo and others are stopping AI tools that log in with user credentials.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.