Saturday, October 4, 2025

☕️ OpenAI flips Sora's copyright policy in 72 hours

Receive a daily summary of what happened in tech, powered by ML and AI.

Thank you! We sent you a verification email.
Oops! Something went wrong while submitting the form.
Join 1,500+ thinkers, builders and investors.

Hi there, this is your daily ☕️ Techpresso.


In today's Techpresso:

🤝 OpenAI flips Sora's copyright policy in 72 hours

🛰️ Bezos predicts gigawatt data centers in space

📈 Google Gemini gains ground on ChatGPT

⚠️ Discord breach leaks user data and photo IDs

🏦 Coinbase applies for a national trust charter

🎁 + 10 other news you might like

🔮 + 3 handpicked research papers and tools

🤝 OpenAI flips Sora's copyright policy in 72 hours LINK

  • After launching with few restrictions and allowing copyrighted material like "South Park," OpenAI quickly blocked these prompts following what appeared to be a forceful response from copyright holders.
  • CEO Sam Altman announced that rights holders will soon get an expanded control system allowing them to set detailed rules for how their specific characters can be generated in Sora.
  • The company also revealed plans for a new revenue model to share money with creators when their content is generated, although concrete details or a timeline have not been released.
  • 🛰️ Bezos predicts gigawatt data centers in space LINK

  • Jeff Bezos predicts that within 20 years, companies will construct enormous gigawatt data centers in space, where constant solar power and easier cooling offer an edge over Earth-based facilities.
  • Uninterrupted by clouds or night, solar panels in Earth orbit could provide consistent energy, while the extreme cold of space offers a straightforward way to cool the powerful server equipment.
  • The project is commercially unfeasible today because launching thousands of metric tons of solar panels and radiators would require well over 150 launches and cost tens of billions of dollars.
  • 📈 Google Gemini gains ground on ChatGPT LINK

  • Google Gemini is gaining ground in the market, with its share of generative AI traffic more than doubling from 6.5 percent to 13.7 percent over the last twelve months.
  • Although ChatGPT still commands the market, its share of generative AI traffic has dropped from 87.1 percent to 73.8 percent, showing a clear decline in its overall dominance.
  • The rest of the market remains fragmented, with competitors like DeepSeek, Perplexity, Grok, Claude, and Microsoft's Copilot each holding less than four percent of the generative AI traffic.
  • ⚠️ Discord breach leaks user data and photo IDs LINK

  • Hackers compromised a third-party customer service provider, exposing Discord users' personally identifying information like real names, usernames, email addresses, and other contact details provided to the support team.
  • The attackers also gained access to photos of government-issued identification documents and partial billing info, which included payment type, purchase history, and the last four credit card digits.
  • The incident also compromised IP addresses, plus the full messages and attachments that people sent to Discord's support agents when they were seeking help with their accounts.
  • 🏦 Coinbase applies for a national trust charter LINK

  • Coinbase has applied for a national trust charter to expand its crypto payment services using stablecoins, while making it clear the company has no intention of becoming a bank.
  • The charter would give Coinbase a single federal overseer instead of a patchwork of state regulators, helping the exchange launch new crypto features much faster across the United States.
  • Other firms like Circle, Ripple, and Paxos are seeking similar national licenses to bypass slow federal lawmaking and the messy landscape of differing state-by-state rules for digital assets.
  • Other news you might like

    Latest research and tools

    fssimu2: a tool that measures the visual similarity between two images, designed to be faster and use less memory than the original reference implementation.LINK

    Characterizing Realistic Workloads on a Commercial Compute-in-SRAM Device: this paper analyzes how a commercial chip that computes inside its memory performs when running realistic, real-world applications.LINK

    Pretraining Under Infinite Compute: this paper finds that for a given amount of computing power, it is better to train a smaller model for more steps than previously thought.LINK


    Want to get the latest news differently? Find us on:

    twitter instagram spotify apple-podcasts


    See you tomorrow for a new dose of ☕️ Techpresso!

    Feeling behind on AI?

    You're not alone. Techpresso is a daily tech newsletter that tracks the latest tech trends and tools you need to know. Join 400,000+ professionals from top companies like OpenAI, Apple, Google and more. 100% FREE.
    Thank you! We sent you a verification email.
    Oops! Something went wrong while submitting the form.
    Join 1,500+ thinkers, builders and investors.
    You're in! Thanks for subscribing to Techpresso :)
    Oops! Something went wrong while submitting the form.
    Join 5,000+ thinkers, builders and investors.