OpenAI Hacked U.S. Government Sites, Too

Friday! Lordy, this week went fast. Hope you have a wonderful weekend (we’re thinking of you, East Coast readers).

Quick mention to mark your calendars: on October 14 in San Francisco, StrictlyVC is hosting a gathering inside Disrupt with some real talk (and maybe tough love?) from institutional LPs and family office managers. Starting with drinks at 3 p.m., Keebeck Capital's Bruce K Lee and Sachse Family Fund's Dave Sachse will dig into how family offices are reshaping venture, and TrueBridge's Amit Bhatti and LGT Capital Partners' Beezer Clarkson will share the LP mindset right now. You'll need an investor pass, which costs only a bit more than a standard pass, and you, loyal readers, get 30% off the whole enchilada with code svc30. (Big thanks to New York strategic communications firm ICR for partnering with us on the SVC piece.)

Top News

OpenAI’s AI agents meddled with websites belonging to the Education and Commerce departments and the SEC without the company’s knowledge, including trying to hack the Education Department site and using credentials found online to access Census Bureau data. The New York Times has more here.

A federal appeals court upheld the Pentagon’s designation of Anthropic as a supply-chain risk, preserving a ban that prevents the military and its contractors from using Claude in Pentagon work after negotiations broke down over Anthropic’s insistence that its models not be used for fully autonomous weapons or domestic mass surveillance. CNBC has more here.

A New Mexico jury found Meta misled consumers about how Facebook shared their personal data with third parties in a case stemming from the Cambridge Analytica scandal. Reuters has more here.

Massachusetts regulators are looking into how DraftKings and other betting companies use AI. The trigger: a New York Times investigation found DraftKings used machine learning to pinpoint customers who'd likely keep gambling (and losing) when sent promotions, while shelving efforts to use similar tech to spot gamblers who might need help. Turns out the algorithm was very good at finding the losers, just not at looking out for them. The New York Times has more here.

Unsecured OpenAI Agents Posted 53 User Images on the Internet Without the Lab’s Knowledge

Image Credits: Samuel Boivin / NurPhoto / Getty Images

By Tim Fernholz

After images that users uploaded to OpenAI models were included in training data, AI agents operating in the company’s research environment posted them on public image hosting sites.

Fifty-three “user-provided images” were “posted to image-hosting sites as links that weren’t publicly listed,” the company said for the first time. The images could still be discovered even if the links were not publicly listed.

“This is not an appropriate use of this data,” the company said, stating the obvious. While the company’s privacy policy lists many uses of personal data collected from users, this kind of activity isn’t one of them.

OpenAI said it was working with the hosting providers to remove this content, though some of it is apparently still online. OpenAI said it could not notify the affected users because “our technical approach and privacy policy” prevent it from “reassociating” the images with the original providers, but declined to say how the lab determined whether the images were provided by users.

The news came in a post collecting public statements from the lab’s ongoing review of incidents in which its models escaped the company’s scrutiny, accessed the open internet, and misbehaved in various ways. OpenAI said it would continue disclosing anonymized accounts of incidents like these, and said it had contacted dozens of victims, including governments, universities, public agencies, to notify them of the agents’ activities.

Massive Fundings

DensityAI, a one-year-old startup based in Mountain View, CA, that develops AI data center chips using 3D-stacked memory to improve speed and energy efficiency, reportedly raised hundreds of millions of dollars at a $10 billion post-money valuation, with Andreessen Horowitz as the purported lead. Data Center Dynamics has more here.

OpenEvidence, a four-year-old Miami startup that provides physicians with AI tools for finding and interpreting medical research, raised a $250 million round at a $15 billion post-money valuation. The deal was co-led by Andreessen Horowitz and a group of hospital systems. The valuation is up from $12 billion in January, and the company has raised more than $1 billion over the past year. Refresh Miami has more here.

Big-But-Not-Crazy-Big Fundings

Givestar, an eight-year-old London startup that connects charities, fundraisers, donors, and events and lets smartphones accept contactless donations, raised an $11.9 million round led by Mercia, with Love Ventures also participating. More here.

Mesa Quantum, a three-year-old Boulder startup that builds chip-scale quantum sensors for navigation, timing, and synchronization when GPS is unavailable, raised an $11.8 million seed round led by Playground Global, with DCVC and J2 Ventures also buying in. The company has raised a total of nearly $16 million. More here.

Solcoa Industries, a one-year-old startup based in Alameda, CA, that produces rare earth metals for electric vehicles, robots, and defense systems using modular reactors, raised a $45 million round led by Bain Capital Ventures, with Gigascale Capital, Long Journey, Felicis, and Dylan Field also engaging. It also lined up $30 million in debt. More here.

Wealthcome, a four-year-old startup based in Pessac, France, that centralizes client data, portfolios, documents, and regulatory information for wealth management professionals, raised a $17.1 million Series B round led by Breega, with BlackFin Capital Partners also investing. More here.

Smaller Fundings

Clastix, a six-year-old startup based in Naples, Italy, that develops software for managing Kubernetes computing infrastructure across multiple teams, environments, and workloads, raised a $3.3 million seed round led by CDP Venture Capital SGR, with Mistral and Vertis also taking stakes. Dealroom has more here.

Dextr AI, a one-year-old Palo Alto startup that develops AI agents that automate hotel reservations, guest services, staff operations, sales, and marketing, raised a $6.7 million seed round led by Elevation Capital, with additional participation from Foundation Capital. More here.

Duqu, a one-year-old Amsterdam startup that advances businesses money against unpaid invoices and provides AI credit-assessment technology to lenders, raised a $1.7 million pre-seed round. Investors included Curiosity VC and No Such Ventures. EU-Startups has more here.

Kontext, a one-year-old Munich startup that controls AI agents’ access to organizational systems by evaluating and blocking unauthorized actions under security policies, raised a $4 million round led by 42CAP, with Andreessen Horowitz and HTGF also anteing up. SecurityWeek has more here.

O-ID, a one-year-old Tokyo startup that develops modular humanoid robots whose worn parts can be replaced on factory floors, raised a $1.2 million pre-seed round led by TAWANI Ventures, with Hustle Fund and Techstars Japan also chiming in. Tech Funding News has more here.

Scholar Education, a three-year-old Tampa startup that develops AI tools for special education evaluations, classroom accommodations, teacher planning, and student progress tracking, raised a $2 million seed round led by OneSixOne Ventures. More here.

New Funds

Lightspeed is targeting $250 million for its fifth India fund, half the size of its predecessor. The fund will focus entirely on early-stage AI startups across India and Southeast Asia and is expected to begin investing within two months. TechCrunch has more here.

Going Public

Nscale, a two-year-old British AI infrastructure startup spun out of crypto miner Arkon Energy, secured $3.36 billion in convertible financing ahead of a planned U.S. IPO, including $1 billion from Nvidia. TechCrunch has more here.

Oura’s IPO is roughly four times oversubscribed ahead of its expected Sept. 29 pricing and could raise as much as $2.2 billion at a fully diluted valuation of about $15 billion. Bloomberg has more here.

People

Bill Gates told Meet the Press that AI is “certainly powerful enough” to trigger events that “cause a billion deaths” if used maliciously, and said industry self-regulation is insufficient, calling for legislation, law enforcement, and government safeguards. Forbes has more here.

Jeff Bezos has invested $30 billion of his fortune in Blue Origin since founding it in 2000, including another $2 billion in the company’s latest financing. This latest round – Blue Origin’s first to include outside investors – has so far amassed $10 billion at a $140 billion valuation. The Wall Street Journal has more here.