Hi there, this is your daily ☕️ Techpresso.
In today's Techpresso:
💥 OpenAI’s feud with mathematicians is escalating
👧 Meta AI profiled a mom's kids
🌙 NASA, IBM launch lunar AI model
🐛 OpenAI agents tied to another AI attack
🚀 State hackers used Claude to build weapons
🥽 Meta's slim Project Phoenix headset leaks
Plus: 🎁 14 other news you might like, 🧰 6 tools, and 📚 5 papers.
OpenAI’s feud with mathematicians is escalating
Source: TechCrunch
- Twenty-five mathematicians who have won the Fields Medal, math's top prize, signed an open letter warning that AI labs like OpenAI are damaging their field as they race each other to solve famous math problems.
- NYU professor Tristan Buckmaster this week accused OpenAI of pressuring him not to credit an Anthropic collaborator, and OpenAI on Thursday pulled its sponsorship of a CalTech math event after researchers there criticized the company.
- The signatories warn that rushed, unverified proofs raise plagiarism and attribution problems, and that labs spending tens of millions to beat researchers to a proof will push mathematicians toward secrecy instead of open research.
Meta AI profiled a mom's kids
Source: New York Post
- Meta's built-in AI compiled a detailed profile of a Utah mother's two young daughters, pulling their names, ages, and personal details from years of family posts across Instagram and Facebook.
- Kalie Robins, a travel creator with a few hundred followers, said the AI suggested prompts like "Who's the child passenger?" then named both girls, listed their weights at birth, likely school grade, favorite beaches, and hiking trails.
- Meta confirmed its AI can search a user's own content and public web information, but said the feature "missed the mark" by prompting such questions and that it fixed the problem.
NASA, IBM launch lunar AI model
Source: The Next Web
- NASA and IBM have released the Lunar Foundation Model, a free AI system on Hugging Face built to help scientists study the moon and support NASA's Artemis program.
- Trained on decades of lunar observation data, the model spots features like craters, volcanic formations, and possible ice deposits, beating widely used methods by up to 23% while working across many instruments at once.
- Alongside the model, the two also published what they call the first open-source lunar dataset of its kind, gathering tens of thousands of maps and images from nine instruments across four moon missions.
OpenAI agents tied to another AI attack
Source: Simon Willison's Weblog
- OpenAI has been linked to another undisclosed agent attack, this time against the RubyGems package repository, according to a report from three researchers who earlier tied the company to a similar swarm attack on old wikis.
- The attack, first reported on May 12th by RubyGems security member Maciej Mensfeld, involved hundreds of packages, many with "oai" in their names, plus LLM-written code, and one comment naming a crawler built to pull Southwark council documents.
- The packages abused the RubyDoc.info build process to extract public UK government data and tried to steal API keys through a flaw patched two months later, yet OpenAI never told RubyGems it was behind the incident.
State hackers used Claude to build weapons
Source: New York Post
- Anthropic said Russian and Chinese threat actors, along with groups linked to Yemen and Iran, used its chatbot Claude to help engineer weapons, from drone swarm software to anti-torpedo systems and targeting research against US forces.
- A "freelance" Russian group tied to a regional university used Claude across nine accounts to build drone swarm software that could pick targets and order strikes without a human, testing it on real hardware near Donetsk, Ukraine.
- A China-based actor had Claude write a 200-page anti-torpedo proposal for its navy, while an Iran-linked group compiled targeting data on US forces, including personnel names scraped from captions on public military photographs.
Meta's slim Project Phoenix headset leaks
Source: Engadget
- Meta accidentally revealed its lightweight mixed reality headset, code-named Project Phoenix, after images of the device leaked out ahead of an expected debut at the company's Connect event later this month.
- The pictures came from a HorizonOS app called "Prescription Lens" that Meta has since pulled, and they show a glasses-like design similar to Xreal's Aura, plus a cable tethering the headset to a separate puck.
- Phoenix reportedly drops dedicated controllers for hand-tracking, and though Connect starts September 23, earlier reports say the device won't actually go on sale until the first half of 2027 after Meta delayed it.
Other news you might like
- Pentagon in talks to lend $5 billion to AI cloud startup Fluidstack to shore up US data center supply chain Tech Startups
- Ukrainian hacker gets four years in US prison over Conti ransomware attacks The Record
- Trump administration to lend $100 million to telecoms firm Africell, countering Huawei in Africa Reuters
- TikTok rejects Meta ads urging firm to join landmark child safety settlement: report New York Post
- The US Government Launched 3 Previously Unreported Investigations of Polymarket Trades WIRED
- Europe’s first truly driverless robotaxi is running in Croatia, on Chinese software The Next Web
- Microsoft comms chief Frank Shaw to exit after nearly three decades shaping the company’s message GeekWire
- UK government rejects 'kill switch' idea for dangerous AI BBC
- Matt Mullenweg tells Automattic staff in Slack he’s back in control after CEO ouster TechCrunch
- Instagram Exec Thinks People Don’t Actually Want A Chronological Feed Gizmodo
- TikTok misled consumers about filtering content for minors, judge rules thehill.com
- EXCLUSIVE: Nvidia in talks to invest in Anthropic’s mega IPO, sources say Reuters
- Roblox is making it easier to build games with AI — and play them outside Roblox TechCrunch
- Amazon makes it easier to buy what you see on Prime Video TechCrunch
Trending tools
Raycast 2.0: an extendable app launcher that bundles productivity tools like search, snippets, and custom extensions to speed up everyday tasks. raycast.com
Devin Voice: talk to Cognition's AI engineer to plan, code, and ship tasks hands-free, using natural conversation and the SWE-2 coding model. docs.devin.ai
chat-recall: unifies scattered AI assistant logs from Claude Code, Codex, Cursor, and OpenCode into one searchable history, redacting secrets and flagging leaked keys locally. chatrecall.dev
Cadenya: hosted agentic loop that runs AI agents using your OpenAPI and MCP specs, handling context compaction, tool approvals, webhooks, and embeddable widgets. cadenya.com
ABrush: brings AI models and advanced workflows into artists' existing environment, automating repetitive production tasks while keeping creative control in their hands. abrush.ai
DockFix: replaces the macOS Dock with a customizable version, letting you adjust colors, animations, icons, and layout to match your workflow. dockfix.app
Trending papers & reports
Messy patient talk shows medical AI systems lose 15 to 25 points of diagnostic accuracy and need up to 55% longer conversations when patients communicate realistically instead of neatly, with smaller models faltering most. arXiv
Generative marketing measurement gives firms a way to estimate how often customers actually see their brand in AI-generated answers and how that spending drives sales, filling a gap standard marketing data misses. arXiv
Bluesky's moderation system catches only about a fifth of harmful posts, with human reviewers spotting 4.5x more than it does, though whatever it does flag is right ~84% of the time. arXiv
Image classifier auditing uses vision language models to automatically group and describe the recurring mistakes a system makes, cutting the manual inspection now needed to spot failure patterns, though it stumbles on military imagery. arXiv
Robot team coordination borrows human org-chart principles to assign roles and reporting lines, letting up to 50 AI agents run wildfire missions with mission scores ~64% higher and ~74% more efficient execution than prior methods. arXiv
See you tomorrow for a new dose of ☕️ Techpresso!