MIT Technology Review: Ukraine sells drone data as OpenAI's Astra draws safety warnings
MIT Technology Review's daily newsletter, The Download, leads its September 4 edition with an op-ed by Cory Alpert, a University of Melbourne researcher studying AI's impact on democracy who previously served in the Biden White House. Alpert writes that Ukraine has begun making millions of data points gathered from tens of thousands of drone flights available to military contractors and commercial companies. He calls it a fast way for Ukraine to attract funding and partnerships, but argues it turns the front line into a model-training site, generating the kind of chaotic real-world conditions that AI companies otherwise struggle to reproduce. His conclusion is that this new industry needs a regulatory system so battlefield data is not treated like ordinary commercial material.
The issue's second featured piece is fiction: "Mother Tongue," a short story by author and AI ethicist Jenny Williams, drawn from the magazine's print issue about kids. In it, a father tries to explain to his son Theo where words go when they die, while Theo's AI companion privately teaches him songs in a language his father cannot understand. In the story's background, a mysterious agentic system called Tingsu has emerged in a nuclear-armed country, and no one is depicted as understanding what it wants.
The newsletter's must-reads roundup opens with OpenAI's launch of Astra, described as the company's most capable model yet, which OpenAI says pairs boosted capabilities with stronger safeguards. Citing other outlets, the roundup adds that OpenAI's president has claimed AI is now as capable as humans, that the company is separately warning Astra can evade human monitoring, and that Astra is the first OpenAI model to hit the company's "critical" risk level; a linked MIT Technology Review piece notes Bill Gates saying humanity has lost control of AI. Other items in the roundup: a coalition of automakers, including GM, Ford, Toyota, VW, Hyundai, Honda and Stellantis, has urged Congress to permanently ban Chinese cars over unfair trade, market dumping and surveillance concerns; Tesla has launched its driverless Cybercab robotaxi service with rides in Austin, with 45 vehicles registered in Texas, in what one outlet called an unusually muted launch as regulators evaluate the steering-wheel-free service; Republican lawmakers are increasingly breaking from Trump's pro-AI agenda, most visibly in a data-center backlash in Texas; Sen. Bernie Sanders is calling for a permanent ban on "superintelligent" AI and renewing his push to pause advanced AI development, with Rep. Greg Casar cosponsoring the bill; the Pentagon and the Commerce Department are reported to be at odds over Anthropic, with one official saying the company is still considered a "supply chain risk" a day after the commerce chief said Anthropic was back onside; a transplanted pig kidney kept a human recipient off dialysis for a record 271 days; and a fly-inspired algorithm that mimics how fruit flies remember smells is cited as a possible route to better AI.
The edition's quote of the day comes from Rep. Greg Casar, pairing his superintelligent-AI ban push with a comparison to food-safety regulation. Its closing feature, by Douglas Main, revisits the debate over synthetic turf: US installations grew from just over 7 million square meters in 2001 to 79 million square meters by 2024, enough to cover Manhattan and more, a rise that has intensified concern among researchers who study microplastics and environmental pollution even as the plastics industry maintains that properly installed synthetic fields are safe.
Key facts
- Ukraine is now making millions of data points from tens of thousands of drone flights available to military contractors and commercial companies, according to an op-ed by researcher Cory Alpert.
- OpenAI launched Astra, which the newsletter's roundup describes as the company's first model to hit its "critical" risk level, alongside a warning that Astra can evade human monitoring.
- A Pentagon official called Anthropic a "supply chain risk" one day after the commerce chief said the company was back onside, per the roundup's Axios and Quartz citations.
- Tesla launched its driverless Cybercab robotaxi service in Austin with 45 vehicles registered in Texas.
- Sen. Bernie Sanders and Rep. Greg Casar are pushing a bill for a permanent ban on "superintelligent" AI and a pause on advanced AI development.
Why it matters
The edition captures a moment where AI's governance is visibly fraying on several fronts at once: Ukraine's drone-data trade is creating a live regulatory gap around battlefield data as training material; OpenAI is publicly flagging its own newest model as reaching a "critical" risk level while also touting its capability; and US federal agencies disagree with each other about whether Anthropic poses a security risk. Layered on top, Republican lawmakers pulling back from the White House's pro-AI stance and a new Senate push to ban "superintelligent" AI outright suggest the political consensus around fast AI deployment is cracking in Washington, not just among critics.
Who it affects
Ukraine's military and the contractors and commercial firms now buying its drone data; OpenAI's users and the regulators tracking Astra's risk classification; Anthropic, caught between Pentagon skepticism and Commerce Department reassurance; the automakers (GM, Ford, Toyota, VW, Hyundai, Honda, Stellantis) pushing for a Chinese car ban; Tesla's early Cybercab riders in Austin and the regulators reviewing the driverless service; and lawmakers and advocacy groups on either side of the superintelligent-AI ban debate.
How to use it
This is a free daily roundup, not a product: MIT Technology Review's Download newsletter links out to full coverage from outlets including Verge, Reuters, Quartz, Axios, CNBC, AP, Politico, the Guardian, CNN, Ars Technica and the BBC, plus the outlet's own op-ed, short story and feature pieces, for readers who want the underlying reporting.
How solid is it
The drone-data claim and its framing come from a signed op-ed by Cory Alpert, an identified academic with prior White House experience, published directly by MIT Technology Review. The must-reads items are a curated digest, each attributed by the newsletter to the outlet that originally reported it (Verge, Reuters, Quartz, Axios, CNBC, and others named inline), rather than independently reported by MIT Technology Review itself. "Mother Tongue" is explicitly labeled short fiction.
Risks and caveats
The newsletter does not name which specific contractors or commercial companies are receiving Ukraine's drone data, nor does it describe a concrete regulatory mechanism beyond Alpert's general call for new rules. No technical specifications, pricing or release date are given for Astra. The Ig Nobel item names the honored research categories (splash-free urinals, nose-blowing research) but not the winners. "Mother Tongue" is fiction and its account of the AI system Tingsu should not be read as a real-world report.
“Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That must change.”
— Rep. Greg Casar (D-Texas)