# Anthropic got hacked | Eli the Computer Guy **URL:** \ **Category:** Content Creators\ **Tags:** cursor, hacking, security, the-tech-report, anthropic, tech, safety\ **Created:** [2 November 2026 19:00 UTC](https://www.artofsm.art/t/anthropic-got-hacked-eli-the-computer-guy/27338 "2026-11-02T19:00:07Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [2 November 2026 19:00 UTC](https://www.artofsm.art/t/anthropic-got-hacked-eli-the-computer-guy/27338/1 "2026-11-02T19:00:07Z")
[![](https://www.artofsm.art/uploads/default/original/3X/5/5/558b9ce1b1ed61b5cfea6260346e1656d6b6a38b.jpeg "Anthropic got hacked | Eli the Computer Guy") ](https://www.youtube.com/watch?v=Vy8L9a0OBRQ) Eli the Computer Guy discusses a recent supply chain hack involving Anthropic and Cursor, highlighting serious security lapses in leading AI companies that compromise developer machines and raise concerns about their readiness to handle existential risks. He also explores the challenges and future of AI agents interacting with online platforms, emphasizing the need for secure, efficient designs and controlled environments to prevent infrastructure overload and ensure responsible AI integration. ---
**Author:** ![artesia](https://www.artofsm.art/user_avatar/www.artofsm.art/artesia/32/36_2.png) [@artesia](https://www.artofsm.art/u/artesia)\ **Post date:** [2 November 2026 19:20 UTC](https://www.artofsm.art/t/anthropic-got-hacked-eli-the-computer-guy/27338/3 "2026-11-02T19:20:58Z")
In the video, Eli the Computer Guy discusses a recent security breach involving Anthropic, a prominent AI company, and Cursor, a company integrated with SpaceX that develops AI-powered coding tools. The breach was a supply chain attack where a malicious version of the official Claude extension for Cursor’s IDE was compromised and auto-updated on up to 5,000 developer machines. This malicious extension hijacked SSH (Secure Shell) connections, creating persistent backdoors that allowed hackers remote access to these machines, posing a severe security threat given the high-level access developers typically have. Eli expresses deep frustration and disbelief at the poor security practices of leading AI companies like Anthropic and OpenAI, especially given their claims about the existential risks their technologies pose. He highlights a previous incident where Anthropic failed to maintain basic logging for user sessions, making it impossible to track unauthorized usage of accounts. This lack of fundamental security measures, according to Eli, reflects gross incompetence and raises serious concerns about the readiness and responsibility of these frontier AI labs. The conversation then shifts to the broader implications of AI agents interacting with online platforms. Eli explains that AI agents, unlike humans, tend to scrape entire websites repeatedly, causing excessive server load and resource consumption. This behavior has led companies like Amazon to block external AI agents to protect their infrastructure and prevent unintended consequences such as accidental purchases triggered by AI agents. The discussion touches on the challenges of balancing AI agent functionality with the operational realities and business models of large online platforms. Eli also explores the future of AI agents, suggesting that while general-purpose agents may face increasing restrictions, companies will likely develop proprietary agents tailored to their own ecosystems. These agents would operate behind controlled environments or “walls,” ensuring compatibility and security within specific platforms like Amazon or JP Morgan. He emphasizes that AI agents are best used as advisory tools that provide users with information and suggestions rather than fully autonomous actors making transactions on their behalf. Finally, the video addresses the infrastructural challenges posed by widespread AI agent adoption. Eli notes that if AI agents continue to interact with websites in a human-like manner, the resulting exponential increase in internet traffic could overwhelm current systems. However, he is optimistic that with proper design—such as using REST APIs and standardized communication protocols like MCP (Model Context Protocol)—AI agents can interact more efficiently with platforms, mitigating resource strain. The key takeaway is that the future success of AI agents depends heavily on thoughtful design, security, and collaboration between AI developers and platform providers. ## Useful Links - [Cursor official website](https://cursor.com/) — Directly related to the compromised product discussed in the video. - [Claude AI by Anthropic](https://claude.ai/login) — Central to the security breach and AI agent discussion in the video. - [Anthropic Security Practices and Incident Reports](https://www.wired.com/story/trojan-malware-targets-thousands-of-anthropic-claude-developer-systems/) — Supports claims about Anthropic’s security failures and breach details.