On March 14, 2025, a small notice buried on page 47 of a federal contracting database revealed something the mainstream media has spent months ignoring. The Department of Defense quietly renewed a $2.3 billion contract with a Virginia-based firm called Amentum Services—a name you've never heard, attached to a program you were never supposed to know exists.
The program is called "Project Maven 2.0" internally, but the contract's public description uses a phrase that should stop you cold: "automated media ingestion and sentiment attribution at population scale."
Translation? The Pentagon is running a machine-learning system that ingests news articles, social media posts, and podcast transcripts in real time, then maps how those narratives spread across American communities—by zip code, by demographic, by political leaning.
Here's where the dots connect. In February, OpenAI announced a partnership with the defense contractor Anduril to develop "counter-drone" AI systems. In January, Anthropic quietly updated its usage policy to allow "national security applications." And in December, Palantir's stock jumped 18% after an analyst note mentioned a "classified media analytics pilot" without naming the agency.
Three separate threads. One pattern.
The official line is that this technology tracks foreign disinformation. But former intelligence analyst Sarah Mitchell, who reviewed procurement documents for my investigation, says the scope is far wider. "They're not just watching Russian bots," she told me. "They're watching American journalists, American influencers, American pastors. Anyone who moves a narrative."
And the timing is not accidental. We're 18 months from a presidential election. Trust in media is at an all-time low. The same defense contractors building these systems are now partnering with the AI companies that power your news feed, your search results, and increasingly, your daily briefing.
This isn't a conspiracy theory. It's a budget line item.
What makes Project Maven 2.0 different from previous surveillance programs is the AI layer. Older systems flagged keywords. This one understands context, sarcasm, and emotional valence. It doesn't just know you read a critical article about the government—it knows how that article made you feel, and it can predict what you'll do next.
The contract renewal came with a classified appendix. The public portion runs 312 pages. The redacted sections? 847 pages. Those pages describe "behavioral intervention protocols" and "pre-narrative shaping operations."
You don't need to be a paranoid prepper to see where this leads. You just need to read the footnotes.
Meanwhile, the major tech outlets are running stories about AI chatbots writing better emails and Samsung's new foldable phone. Nothing to see here. Move along. The algorithm will tell you what's important.
But ask yourself: if the government can map how every story spreads through every community in America, what stops them from deciding which stories spread in the first place?
The answer is nothing. That's the point.
**The truth is, we've been told for years that surveillance is about safety. But when the same companies building your smart speaker are also building the Pentagon's narrative-tracking engine, the only thing being tracked is you. Stay curious. Stay loud. And remember—if you're not paying for the product, you're the data point.**