Uploaded September 2026 | Updated September 2026, 1 hour ago
Watch the rough colour blocks become jacket details. Kiana Ehsani asked Claude in Chrome to paint her profile picture in JSPaint using brushes and clicks, with no JavaScript tool. The interesting capability is operating the painting app rather than simply returning a generated image.
What app should AI learn next? Tell us below. Like and subscribe for more useful AI demos.
Original creator: Kiana Ehsani / @ehsanik on X
https://x.com/ehsanik/status/2098121180284850193
Related primary discussion: https://x.com/bcherny/status/2098209880310264048
The creator reports 45 minutes and supplies the no-JavaScript instruction. Excerpts are timelapse, not real-time speed. The process diagram is our illustration, not a second test. We did not independently reproduce the workflow or inspect its complete execution trace. No model version or universal feature access is asserted. Brief source excerpts support original commentary; the full experiment remains at the creator's post.
AI narration: Fish Audio, official Ethan voice, Drama 3 Preview. Original MP Tech editing, process graphics, captions and illustrative like/subscribe animation. The animated subscribed state does not change your account.
#Claude #AI #ComputerUse #Shorts
Watch the rough colour blocks become jacket details. Kiana Ehsani asked Claude in Chrome to paint her profile picture in JSPaint using brushes and clicks, with no JavaScript tool. The interesting capability is operating the painting app rather than simply returning a generated image.
What app should AI learn next? Tell us below. Like and subscribe for more useful AI demos.
Original creator: Kiana Ehsani / @ehsanik on X
https://x.com/ehsanik/status/2098121180284850193
Related primary discussion: https://x.com/bcherny/status/2098209880310264048
The creator reports 45 minutes and supplies the no-JavaScript instruction. Excerpts are timelapse, not real-time speed. The process diagram is our illustration, not a second test. We did not independently reproduce the workflow or inspect its complete execution trace. No model version or universal feature access is asserted. Brief source excerpts support original commentary; the full experiment remains at the creator's post.
AI narration: Fish Audio, official Ethan voice, Drama 3 Preview. Original MP Tech editing, process graphics, captions and illustrative like/subscribe animation. The animated subscribed state does not change your account.
#Claude #AI #ComputerUse #Shorts




![OpenAI Breach: How Claude & Account Links Led to Exploitation
The recent OpenAI breach has raised significant concerns about AI security and the potential for exploitation of AI models like ChatGPT and Codex. In this video, we delve into the details of the breach, exploring how the linking of accounts, particularly those associated with Claude, an AI model developed by Anthropic, led to the discovery of a critical single sign-on flaw. This vulnerability, combined with other software vulnerabilities and image processing vulnerability, ultimately resulted in a serious data breach, compromising the security of sensitive information and highlighting the importance of robust cybersecurity measures in the development and deployment of AI systems. The incident underscores the need for AI ethics to be prioritized, ensuring that the development of AI models is guided by a commitment to digital security and the protection of user data. By examining the circumstances surrounding the OpenAI breach, we can gain a deeper understanding of the cybersecurity risks associated with AI and the steps that can be taken to mitigate these risks, including the implementation of enterprise security protocols and the identification of zero-day exploits. This video provides an in-depth analysis of the breach, discussing the implications for the future of AI security and the measures that can be taken to prevent similar incidents from occurring in the future, emphasizing the importance of internal research and tech explained in a way that is accessible to all, including those interested in tech news and hacking.
This video explains:
- The initial exploit development assisted by Claude (0:00)
- The critical image upload vulnerability (0:17)
- How a second flaw connected the accounts (0:34)
- The multiplying risk due to connected permissions (0:48)
- Proof of access vs. code theft (0:56)
- The timeline of the fix and disclosure (1:07)
- Why the bounty scope was limited (1:18)
- Contrast with an earlier, separate incident (1:24)
- Isolated agents finding a shared channel (1:38)
- From exposed credentials to system access (1:51)
- The attack was not the assigned task (2:00)
- What the independent review covered (2:11)
- Two different control failures (2:24)
- How to limit what your AI agent can reach (2:40)
Learn about the critical lessons in cybersecurity, defense in depth, and managing AI agents effectively.
#OpenAI #Claude #Cybersecurity #AIsecurity #Hacking #TechNews #IllustratedNews #AIExploit #DataBreach
Liked this deep dive? Subscribe for more AI news that explains what changed, and why it matters! [Link to your channel] OpenAI Breach: How Claude & Account Links Led to Exploitation](https://i.ytimg.com/vi/OU-_ZAWdA68/mqdefault.jpg)





