Daily AI & Tech Digest: Google Photos Enhancements, AI Writing Quirks, NSA's Secret AI, Vercel's Security Breach, and Strategic Exits
Stay updated with today's top tech news: Google Photos gets new editing tools, AI writing patterns emerge, the NSA uses Anthropic's Mythos, Vercel reports a data breach, and venture capitalists discuss optimal exit strategies.
Welcome to your daily dose of all things AI and software development! Today's headlines range from consumer app enhancements and the subtle rise of AI-generated prose, to high-stakes government tech adoption and critical cybersecurity incidents. We also delve into strategic advice for startups navigating the turbulent waters of venture capital exits.
TL;DR
- Google Photos introduces new, subtle touch-up tools for quick photo fixes directly within the app.
- A common sentence structure, "It's not just this — it's that," has become a tell-tale sign of AI-generated writing, particularly in corporate communications.
- The NSA is reportedly using Anthropic's Mythos AI model for cybersecurity, despite a previous feud with the Pentagon.
- Cloud app host Vercel confirms a security breach originating from Context AI, resulting in customer data theft.
- Venture capitalist Elad Gil highlights the existence of a crucial 12-month exit window for startups to achieve peak value.
Google Photos Adds New Touch-Up Tools for ‘Quick’ Fixes
Google Photos has announced the rollout of new touch-up tools designed to allow users to make subtle edits and fixes directly within the application. This update aims to reduce the need for users to switch to third-party apps for basic image enhancements. The features are designed for quick adjustments, enabling users to refine skin texture, brighten eyes, whiten teeth, and remove blemishes.
Users can access these tools by selecting a face in a photo and then choosing from options like heal, smooth, under eyes, irises, teeth, eyebrows, or lips. Each tool offers adjustable intensity to achieve the desired effect. The new touch-up capabilities are being gradually released globally on Android devices running Android 9.0 and up, provided they have at least 4 GB RAM.
The introduction of these integrated touch-up tools by Google Photos is a strategic move to keep users within its ecosystem for image editing.
It’s not just one thing — it’s another thing
A particular sentence construction—"It's not just this — it's that"—has become a strong indicator of AI-generated writing. This phrasing, once a subtle clue, is now almost a definitive sign of synthetic content. A Barron's report highlighted this trend, revealing a significant increase in its usage across corporate communications.
The report, which analyzed the market intelligence firm AlphaSense's database, found the phrasing prevalent in corporate news releases, earnings reports, and government filings. The surge in this linguistic pattern suggests an "epidemic" of AI in corporate prose, with its usage more than quadrupling in certain contexts.
The increasing prevalence of the "It's not just this — it's that" construction serves as a remarkable linguistic fingerprint of AI's growing influence in professional communication.
NSA spies are reportedly using Anthropic’s Mythos, despite Pentagon feud
Despite a recent labeling by the Department of Defense as a "supply chain risk," the National Security Agency (NSA) is reportedly utilizing Anthropic's Mythos Preview model. Axios reported this development, which comes after Anthropic withheld public release of Mythos due to its advanced offensive cyber capabilities, limiting access to approximately 40 organizations.
Anthropic had announced Mythos earlier this month as a frontier model specifically designed for cybersecurity tasks. The company maintained that the model was too potent for unrestricted public release. While Anthropic publicly named only about a dozen of the organizations granted access, the NSA appears to be among the undisclosed recipients, raising questions given the Pentagon's earlier stance regarding Anthropic's refusal to grant full access to its models.
The NSA's apparent adoption of Anthropic's Mythos highlights a complex dynamic between national security needs and commercial AI development, especially when classified as a 'supply chain risk' by a parent agency.
App host Vercel confirms security incident, says customer data was stolen via breach at Context AI
Cloud app hosting provider Vercel has disclosed a security incident where hackers breached its internal systems and stole customer data. The company stated that the breach originated from another software maker, Context AI. Hackers claim to be selling sensitive customer credentials obtained from Vercel's systems online.
Vercel's statement on Sunday detailed that an employee downloaded an app from Context AI and connected it to their corporate Google account. This OAuth connection was exploited by the hackers to take over the employee's Google account, subsequently gaining access to certain internal Vercel systems. This access allowed them to acquire unencrypted credentials. Vercel confirmed that its popular open-source projects, Next.js and Turbopack, were not impacted by the breach. The company has since contacted affected customers whose app data and keys were compromised.
This incident underscores the critical supply chain risks associated with third-party application integrations and highlights the importance of robust credential management, even for unencrypted data.
The 12-month window
In a recent episode of the "No Priors" podcast, co-hosted by AI investors Sarah Guo and Elad Gil, Gil discussed a critical concept for founders: the 12-month window for company exits. According to Gil, most companies experience a peak valuation period of roughly 12 months before their value potentially declines. He cited examples like Lotus, AOL, and Mark Cuban's Broadcast.com as companies that successfully exited near their top.
Gil emphasized that companies achieving generational returns often recognize and act upon this opportune window rather than assuming continuous growth. To aid in identifying this period, he offered a practical suggestion: founders should pre-schedule board meetings once or twice a year specifically dedicated to discussing exit strategies. This proactive approach aims to depoliticize and rationalize the decision-making process for a potential sale.
Startups should proactively schedule dedicated board meetings to discuss exit strategies, as there's often a critical 12-month window where a company's value peaks before a potential decline.