Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 🖼️ Newest ChatGPT image trend

• 🧑‍💻 Wikipedia gives away data to AI developers

• 🧠 Google’s cheapest thinking AI

• 📺 Watch Sam Altman at TED 2025

Let’s get right into it!

Sign Up | Advertise | Tools

STARTUPS

Newest ChatGPT image trend

A new trend is emerging where users are leveraging ChatGPT’s latest models, o3 and o4-mini, to identify locations from photos, raising privacy concerns. These models can analyze and enhance even blurry images, allowing users to accurately guess cities, landmarks, or venues—sometimes outperforming older models like GPT-4o.

While the capability can be useful for accessibility or emergency response, it also opens the door to potential misuse, such as doxxing people from social media photos. Despite OpenAI’s safeguards to block sensitive info and prevent identifying private individuals, critics note the lack of clear restrictions against this kind of “reverse location lookup.”

STARTUPS

Wikipedia gives away data to AI developers

Wikipedia is trying to reduce the strain caused by AI bots scraping its content by releasing a machine-learning-friendly dataset in partnership with Kaggle. This new beta dataset, available in English and French, offers structured, machine-readable Wikipedia content—such as summaries, infoboxes, and article sections (excluding references and media files)—and is openly licensed.

Designed to support AI training, fine-tuning, and analysis, it offers a cleaner, more efficient alternative to scraping raw text and aims to help smaller AI developers and researchers access Wikipedia data without overloading its servers.

BIG TECH

Google’s cheapest thinking AI

Google has introduced Gemini 2.5 Flash, a cost-efficient AI model designed for developers and enterprises that offers controllable reasoning capabilities for different tasks. Unlike models meant for general users, Gemini 2.5 Flash allows users to set "thinking budgets" and toggle reasoning on or off depending on task complexity, helping optimize performance and cost.

Priced competitively at $0.15–$0.60 per million tokens (and up to $3.50 with reasoning), it’s Google’s lowest-latency model yet, performing exceptionally well in benchmarks—second only to Gemini 2.5 Pro and rivaling OpenAI’s o4-mini in some tests. It’s currently available for preview via the Gemini API, Google AI Studio, and Vertex AI.

STARTUPS

Watch Sam Altman at TED 2025

At TED 2025, OpenAI CEO Sam Altman discussed the company’s explosive growth to 800 million weekly users, the infrastructure challenges caused by high demand, and the growing scrutiny surrounding AI’s societal impact. He acknowledged OpenAI’s evolution from a nonprofit to a $300 billion tech giant and addressed criticisms about power consolidation and safety risks, especially with autonomous AI agents.

Altman announced efforts to compensate artists for AI-generated work and hinted at loosening content moderation to reflect broader user preferences. Despite lacking clear answers on AGI definitions and safeguards, Altman emphasized OpenAI’s commitment to building safe, beneficial AI amid rising legal, ethical, and operational pressures.