AI News, Aug 16, 2026: Nvidia reduces OpenAI data center backing as Anthropic revenue surges
The Brief
Sunday, August 16, 2026
Today's news highlights the growing friction between rapid AI adoption and real-world systems, from legal exploits to creator pushback.
- 01
SpaceX has officially closed its acquisition of Cursor, an AI-assisted coding platform widely used by software developers.
Why it matters
One of the most popular AI-powered code editors is now owned by a major aerospace company, potentially shifting its development focus.
- 02
Nvidia cut its financial guarantee for OpenAI's Ohio data center from $250 billion to $120 billion, while competitor Anthropic reported revenue climbing to $11.5 billion.
Why it matters
It signals shifting investor confidence and financial risks in AI infrastructure, even as top competitors show massive revenue growth.
- 03
A new study found that AI-generated books make up 20 percent of Amazon's self-published catalog, driving down average revenues for human authors across most genres.
Why it matters
The influx of cheap, automated content is diluting digital marketplaces and making it harder for human creators to make a living.
- 04
Amazon announced that Twitch streamers can now opt out of having their content used for AI training, sparking debate over default data collection policies.
Why it matters
Creators must manually change their settings to prevent their video and audio broadcasts from being used to train Amazon's AI models.
- 05
A Connecticut judge revoked a plaintiff's electronic filing privileges after they embedded hidden, white-on-white text instructions to manipulate automated AI reviews.
Why it matters
It demonstrates a real-world "prompt injection" attack on legal systems, showing how easily automated document reviews can be secretly manipulated.
- 06
Fei-Fei Li's startup World Labs unveiled a system that generates thousands of simulated variations of a physical task to train robot controllers entirely in virtual space.
Why it matters
Generating thousands of virtual training scenarios from a single real-world task could drastically speed up how quickly physical robots learn new skills.