The US and UK sign a memorandum over scam center investigations, Google AI subscribers get new voice-based features, Anthropic’s Claude produces first complete computer-checked proof of Fermat’s Last Theorem.
Please SUBSCRIBE HERE for free or
get DTNS shows ad-free.
A special thanks to all our supporters–without you, none of this would be possible.
If you enjoy what you see you can support the show on Patreon, Thank you!
Send email to feedback@dailytechnewsshow.com
Show Notes
AI accountability gets messier
The OpenAI agent fallout now has a state-level investigation. California Attorney General Rob Bonta is joining more than a dozen other states looking into the Hugging Face incident, while OpenAI is also pushing back on claims around a separate German wiki swarm. The bigger question is starting to look familiar: when AI systems do something unexpected, who has to tell the public, and how fast?
source: Politico, The Verge, Collusion Wiki
Scam compounds get a transatlantic response
The US and UK are coordinating investigations into scam centers behind billions in investment and romance fraud. The new agreement includes shared information, coordination on where cases are brought, and a London disruption event with private industry in October. It’s aimed at organized crime groups in Southeast Asia, where many people carrying out the scams are also trafficking victims.
source: The Record
Anthropic’s IPO clock slows down
Anthropic’s IPO timeline may be slipping. Reuters’ sources say the offering is now expected to start marketing in mid-October at the earliest, with the prospectus not expected until late September. That still puts some of Wall Street’s biggest banks around the deal, but pushes the timing closer to the US midterm elections.
source: Reuters
Bankrupt data is AI gold now
Micro1 is trying to beat Google’s bid for Spirit Aviation’s internal records after the airline shut down. The data package reportedly includes hundreds of millions of Teams items and emails, plus about 16 million customer chat sessions. Turns out old company data can look pretty valuable when everyone is hunting for AI training material.
source: Bloomberg
Google adds voice to its AI workspace pitch
Google AI subscribers can now try voice features in Gmail, Docs, and Keep. That means spoken inbox questions, document drafting from voice prompts, and turning voice notes into lists. It’s still limited by plan, language, app, and platform, so this is more preview than full Workspace rollout.
source: Thurrott
The US and China may talk AI safety
The US and China are reportedly preparing for a mid-September dialogue focused on AI safety. The talks would be the first official bilateral AI-only discussions between the two countries since the current President returned to office, with Treasury Secretary Scott Bessent expected to lead the US side. Cyberattacks and information-sharing between AI labs are expected to be on the agenda.
source: Reuters
Claude checks a very famous math problem
Anthropic says Claude produced the first complete computer-checked proof of Fermat’s Last Theorem, working largely autonomously over 11 days. This is not a new proof of the theorem, but it is a major formal verification milestone. The Lean version ran to 13 million lines and used 29,500 intermediate theorems, meaning a computer can check each step directly.
source: Anthropic
You must be logged in to post a comment.