Back to articles
DevPickView feed
Hacker NewsHN submitter: armcat
Original published: Updated on DevPick:

Do the math: How we took over ChatGPT accounts

AI summary
Editorial policyUpdated on DevPick:

Researchers bypassed ChatGPT's sandbox and content security policy, enabling them to take over any user's account simply by sending a link.

Key points

  • •The research team developed a method to bypass ChatGPT's sandbox.
  • •They successfully circumvented the Content Security Policy (CSP).
  • •Account takeover was possible by merely sending a link to the victim.
Read the original(opens in a new tab)
#Security & Misuse#Major Players#ChatGPT

Related articles

See more
PC Watch

An AI secretary that communicates via "that matter". The essence of ChatGPT's "dots" is not always-on operation

The new feature "dots" added to ChatGPT at the end of September 2026 is available only on Pro plans or higher, emphasizing shared context over always-on operation.

PC Watch

ChatGPT Image Generation 50% Faster, Now Supports GIF Animation

OpenAI announced 'ChatGPT Images 2.6', which is 50% faster and now supports GIF animation.

Hacker News

Ask HN: What's your current AI agentic setup?

A Hacker News thread soliciting user input on their current daily AI agentic setups, including the specific AI subscriptions they pay for, their usage tiers, and whether they utilize models outside of those subscriptions.

The Verge

OpenAI says ChatGPT is catching up to Google in search

OpenAI announced that search queries via ChatGPT now equal 20 percent of the volume run through Google, reflecting a shift in user behavior.

TechCrunch

OpenAI is finally merging ChatGPT’s Chat and Work tabs

OpenAI announced the integration of ChatGPT's chat and Work tabs, where ChatGPT and Codex will now draw usage from the same allocation pool.