AI news digest, March 15-18, 2026

Anthropic fights the Pentagon in court, Mistral unveils Forge for training custom models, employers start counting employee tokens.

Author: Michael Kokin ·

Anthropic is fighting the Pentagon in court, and unexpectedly the entire industry — including competitors — has sided with them. Mistral unveiled Forge, which could change corporate AI: companies can now train models from scratch instead of fine-tuning others'. And employers started counting employee tokens — AI ROI is becoming a real metric (finally!).

Microsoft considers lawsuit against Amazon and OpenAI
Microsoft is exploring legal action over AWS potentially offering OpenAI Frontier in breach of its contract with Microsoft.
-> read

OpenAI released GPT-5.4 mini and nano
Mini and nano versions of the flagship GPT-5.4 are now available — AI is getting cheaper and faster for mass adoption.

US DOJ: Anthropic can't be trusted with military systems
In response to Anthropic's lawsuit, the government stated the company was lawfully penalized for trying to restrict military use of Claude.

OpenAI's mental health experts unanimously opposed "frivolous" ChatGPT
OpenAI's internal psychology team opposed launching an adult content mode, but leadership decided otherwise.

AI drone startup Swarmer soared 520% on Nasdaq debut
The drone software company valued at $380M+ after IPO — its tech has already been tested by the military.

Companies started tracking employee AI spending
Employers are counting tokens used to measure AI ROI — a new productivity metric.

xAI employees complain about chaos at the company
Constant reorganizations at Musk's company are destroying team morale.

Mistral bets on "build your own AI"
Mistral Forge lets companies train their own models from scratch on proprietary data — a challenge to OpenAI and Anthropic in the enterprise market.