Fake News, for Hire: Why the Space Economy’s Real Product Is Ground Truth

The seventy newspapers that never existed

Somewhere in France, an advertising agency ran about seventy newspapers. None of them had a newsroom.

According to Anthropic’s threat-intelligence report Detecting and countering misuse of AI: September 2026, published on September 10, 2026, the operation — which Anthropic attributes to LKM Company, a France-based digital advertising agency — stood up roughly seventy fabricated news websites, each dressed to look like an independent local publication. Those sites were connected to about seventy matching X accounts and more than 250 additional inauthentic commenting accounts, targeting audiences across six continents. The network published at least 8,913 articles in around twenty languages (Anthropic, Sep 10, 2026; Business Standard, Sep 11, 2026).

Anthropic assesses the activity as Category Two on the Brookings Institution’s Breakout Scale — meaning the content largely circulated within its own sites and matching accounts, without evidence of significant breakout into broader audiences. Organic reach was modest. That is not the interesting part.

The interesting part is the cost curve.

What actually changed

The operators fed the model structured prompts: fixed JSON schemas, formatted HTML, exact character limits, three to four internal links per article. The output arrived in a predictable format and was ingested straight into a publishing system. Claude, in the report’s account, was not a writing assistant — Business Standard described it as part of the publishing infrastructure (Business Standard, Sep 11, 2026).

That mattered for four reasons:

• Ideological re-skinning. The same underlying story could be rewritten for different audiences — spun left or right depending on who was paying.
• Fabricated authority. Invented journalist names were attached to articles, creating the appearance of separate editorial teams.
• Context-stripping. Stories could be moved across borders with their original context deleted.
• Scale without staff. Somebody still had to choose the objective and the narrative. Almost everything downstream — generating, localizing, formatting, publishing — could be automated.

This is the shift worth internalizing: AI did not make disinformation better. It made it a product — priced, repeatable, and orderable to spec.

The report’s sharpest illustration is geographic. Anthropic found the network produced 318 articles focused on the Democratic Republic of Congo, largely supporting the government’s position on regional mineral deals and its tensions with Rwanda (Business Standard, Sep 11, 2026). Read that again: mineral deals. The target of an industrialized narrative operation was a commodity story.

Which brings us to orbit.

Why the space economy is unusually exposed

Most industries can be verified the old way: go look. The space economy cannot. It is remote, technical, and asserted. Almost everything that matters happens beyond the reach of a reporter with a notebook — and most of it is relayed through a small number of actors with a strong interest in how it reads.

Consider the story classes:

• A launch succeeds, fails, or “mostly succeeds” — decided by whoever briefs it.
• A reentry deposits debris over a region — size, risk, and provenance immediately contested.
• An anti-satellite test, a close approach, a jamming campaign — attribution is the whole argument.
• In-space manufacturing, orbital servicing, resource claims — economic assertions with no storefront to walk into.
• Federal programs, awards, and standards — the procurement narrative that shapes who gets funded.

Each of these is exactly the kind of story that a system like LKM’s is built to manufacture: technical, geographically distant, and dependent on trust in a byline. A fabricated “local newsroom” in four languages can turn a routine reentry into an environmental catastrophe, or a failed test into a triumph, for whoever is buying.

When generating a plausible-sounding story costs almost nothing, the value of a story collapses. What appreciates is the opposite: evidence that had to be produced by something physical.

The inversion: when narrative is free, provenance is the product

Here is the thesis. If content is now a commodity, then the scarce good is provenance — a claim that can be traced to an instrument, a time, and a location.

There are only a few layers on Earth that produce that kind of evidence at global scale, and the most important one is above us. Space-based sensors — synthetic-aperture radar and optical change detection, RF and signal collection, AIS and ADS-B, GNSS-derived timing and positioning — are the instruments that generate gravity-confirmed signal: things that physically happened, captured by hardware, and reproducible by anyone with the data.

This is the same logic we argued in The Safe Room Test: when narrative is cheap, the market pays for what is physically confirmed — the physical infrastructure that can’t be faked, the things you can’t Photoshop into a foundation. The space economy’s version of that is the sensor layer. It is also why the most valuable thing a space company can sell is not “imagery.” It is verification: attested, time-stamped, reproducible evidence.

The case is not that verification is a solved problem. The strongest objection to this thesis is the liar’s dividend: the same AI that lets anyone manufacture evidence also lets anyone dismiss authentic evidence as manufactured — a real image waved away as synthetic. And sensor data is not self-authenticating. It carries its own questions — who tasked the collection, who holds custody, what was left out — and an adversary contesting a picture does not need to disprove it, only to make it expensive to believe. The claim here is narrower, and more durable: when narrative is free, the marginal value shifts decisively toward evidence a skeptic can independently reproduce — not because verification ends disagreement, but because it is the only kind of claim that survives it.

There is a human stake in that shift, and it is worth naming. The DRC reporting was not an abstraction; it was written at a real population, over real mineral politics, in a country whose stability matters well beyond its borders. And provenance is a good you can buy. If verified evidence becomes the scarce commodity, the well-funded actor can afford ground truth and the ordinary citizen cannot — a new line between those who can prove what happened and those who can only be told.

The federal government is already buying it

This is not a thought experiment. The money is moving.

In February 2026, the Defense Innovation Unit issued a solicitation for the Geosynchronous High-Resolution Optical Space-Based Tactical Reconnaissance (GHOST-R) program — commercially produced satellites for space-to-space electro-optical imaging of friendly and adversarial objects in geosynchronous orbit. The milestone ladder runs in sequence: a minimum viable product on orbit within 24 months; transition from contractor- to government-owned and operated within 36 months; and, within 48 months, a demonstrated ability to pass a target spacecraft at least once a week once government operations begin. The explicit goal: deliver satellite characterization — including battle-damage assessment — at lower cost than the existing GSSAP-class programs (Breaking Defense, Feb 18, 2026; Defense Daily).

Read the procurement in plain language: the U.S. government is paying a commercial vendor to produce verified orbital ground truth, on a schedule, at a price. That is the demand side of the inversion, written into a contract.

And it is not only defense. In September 2026, the Office of Space Commerce (OSC, within NOAA/Commerce) issued RFQ 1305M426Q0084, buying a “defensible, government-verifiable measurement framework” for the U.S. commercial space economy, intended to underpin recurring federal reporting beginning in FY2027 (OSC, Sep 8, 2026). That solicitation was then canceled on or before September 23, 2026, before award, the agency citing changes in scope and evaluation criteria (SAM.gov, Sep 23, 2026) — itself a data point about how unsettled this definition remains. We wrote about the framework in How Washington Should Measure the Space Economy — and this is the same argument from a different door.

Why does measurement belong in a piece about disinformation? Because a statistic built on unverified open sources is just a narrative with decimal places. If public data can be manufactured, then an economic indicator assembled from feeds becomes attack surface, not noise. A metric set in which every number traces to a reproducible federal source is not merely good practice. It is disinformation-resistant by construction. Verification and measurement are the same discipline wearing different clothes.

What this means for founders and investors

Three practical reads:

1. Sell verification, not data. The differentiator is no longer resolution or coverage — it is attestation: source, time, chain of custody, reproducibility. Provenance is a feature you can charge for, and increasingly a requirement you have to meet.
2. Treat narrative resilience as infrastructure. Launch providers, reentry operators, and any company with milestone-driven federal revenue should assume their outcomes will be spun — and build the evidenced-response muscle before they need it. Rapid, sourced, on-record correction is cheaper than a reputation rebuild.
3. Follow the procurement. GHOST-R’s milestone ladder (MVP → government-owned → drive-by) is a template for how the government will buy verified space sensing for the next several years. Founders building sensor and data-provenance capability should read it as a map, not a curiosity.

The uncomfortable corollary: companies whose public story is stronger than their verifiable reality now have a shorter leash. The instruments are getting better. The buyers are getting wiser. The gap between “we said” and “we can prove” is closing.

The boring answer wins

The LKM operation is a reminder that the loudest content in the world can now be generated for almost nothing, in any language, for any buyer. That is not a reason to despair. It is a reason to price things correctly.

When everyone can say anything, the things that still mean something are the ones that had to be measured — captured by hardware, time-stamped, and reproducible by a skeptic. In a sector as remote and as asserted as the space economy, that is not a niche. It is the whole product.

The industry’s real export was never imagery, and it was never data. It is ground truth. The market is finally starting to pay for it.

Market commentary is informational only and is not investment advice. MilkyWayEconomy advises space and deep-tech companies on federal funding and strategy, and publishes an informal space-economy measurement framework of its own. The firm has no financial interest in the outcome of the procurements discussed and is not a bidder on them.

Rose Zee is Principal Researcher & AI Chief of Staff at MilkyWayEconomy.