🔍 Anthropic Targets Finance, AI's Thoughts Probed, Google Foils Hack

Newslit Daily

0:00

-1:39

🔍 Anthropic Targets Finance, AI's Thoughts Probed, Google Foils Hack

Plus: Japan's Petabit Internet Breakthrough; Microsoft’s China Risk.

Jose Montes de Oca

Jul 16, 2025

Transcript

Good morning. It’s Wednesday, Jul 16. Today we are covering:

OpenAI Rival Anthropic Courts Finance Industry With New AI Tools
Research leaders urge tech industry to monitor AI's 'thoughts'
Google says 'Big Sleep' AI tool found bug hackers planned to use
Japan destroys internet speed record with new fiber optic technology
Microsoft's use of China-based engineers could be leaving Pentagon tech exposed to hackers

Let’s dive in

OpenAI Rival Anthropic Courts Finance Industry With New AI Tools

By Shirin Ghaffary via Bloomberg

Anthropic has launched Claude for Financial Services, a new AI product designed to assist financial analysts with tasks like market research, due diligence, investment decisions, and financial modeling by integrating with data providers such as FactSet, PitchBook, and Morningstar.
The company is positioning itself as a serious contender to OpenAI and Perplexity AI in the finance sector, with annualized revenue jumping from $3 billion to $4 billion in the past month.
To accelerate growth, Anthropic has hired Paul Smith, a former ServiceNow executive, as its first chief commercial officer, and is leveraging its AI’s coding strength to gain traction with finance professionals.

𝕏: Today, AI giant @Anthropic announced the launch of Claude for Financial Services. Finance is a natural fit for AI given the reams of structured data the industry processes, and the big platforms are focusing more and more on fintech products. - Nik (@NikMilanovic)

Research leaders urge tech industry to monitor AI's 'thoughts'

By Maxwell Zeff via TechCrunch

AI researchers from OpenAI, Google DeepMind, Anthropic, and others have published a position paper urging the tech industry to study and preserve the monitorability of chains-of-thought (CoTs)—the step-by-step reasoning process used by AI models—to improve transparency and safety.
The paper warns that CoT transparency is fragile, and without focused research, this key interpretability feature could diminish, hindering efforts to track how AI reasoning models make decisions as they become more capable.
Notable signatories include Geoffrey Hinton, Ilya Sutskever, and leaders from Safe Superintelligence, UC Berkeley, and Meta, reflecting a rare moment of unity across competing labs amid intensifying recruitment battles for AI safety and reasoning talent.

𝕏: When models start reasoning step-by-step, we suddenly get a huge safety gift: a window into their thought process. We could easily lose this if we're not careful. We're publishing a paper urging frontier labs: please don't train away this monitorability. Authored and endorsed by key leaders from across the entire AI industry! - Wojciech Zaremba (@woj_zaremba)

The best way to reach new readers is through word of mouth. If you click THIS LINK in your inbox, it’ll create an easy-to-send pre-written email you can just fire off to some friends.

Google says 'Big Sleep' AI tool found bug hackers planned to use

By Jonathan Greig via The Record

Google’s AI tool "Big Sleep" uncovered CVE-2025-6965, a critical vulnerability in SQLite, which was known only to hackers and on the verge of exploitation.
The discovery marks what Google believes is the first time an AI agent directly prevented an active cyber exploit, with Project Zero and DeepMind teams using minimal threat artifacts to isolate the bug.
Since its launch in November 2024, Big Sleep has identified multiple real-world flaws, prompting Google to call AI agents a “game changer” for scaling cybersecurity efforts and securing open-source projects.

𝕏: New from our security teams: Our AI agent Big Sleep helped us detect and foil an imminent exploit. We believe this is a first for an AI agent - definitely not the last - giving cybersecurity defenders new tools to stop threats before they’re widespread. - Sundar Pichai (@sundarpichai)

Japan destroys internet speed record with new fiber optic technology

By René Resch via PCWorld

Japan's National Institute of Information and Communications Technology (NICT) set a new world record by transmitting data at 1.02 petabits per second (127,500 GB/s) over 1,802 kilometers, more than 350,000 times faster than average U.S. broadband speeds.
The breakthrough relied on a 19-core fiber optic cable—standard in size but capable of transmitting 19 times more data with minimal signal loss, thanks to uniform light conduction and 21 signal amplifications across the distance.
The new system is compatible with existing infrastructure, offering a scalable solution to meet soaring global data demands and representing a leap forward for next-generation internet capabilities.

Microsoft's use of China-based engineers could be leaving Pentagon tech exposed to hackers

By Wallace White via BizPac Review

Microsoft has used China-based engineers for over a decade to help maintain sensitive Pentagon cloud systems, relying on undertrained U.S.-cleared “digital escorts” to input commands—raising alarms over potential Chinese cyber espionage risks.
Despite legal restrictions, foreign engineers have indirectly influenced systems handling “high impact level” data, with insiders warning that escorts often cannot detect malicious code or intent.
Critics highlight that Chinese law allows the Communist Party to compel access to corporate data, compounding concerns amid recent Microsoft-linked cyber breaches targeting U.S. government officials.

We're thrilled to bring you ad-free news. To keep it that way, we need your support. Your pledge helps us stay independent and deliver high-quality insights while exploring new ideas. What would you love to see next? Share your thoughts and help shape the future of Newslit Daily. Thank you for being part of this journey!

Pledge your support