AI Security — week of 2026-06-01

June 1, 2026 · 29 developments

ChatGPT Vulnerable to Phishing Attacks

The Hacker News · 2026-05-29 · corroborated · reporting

ChatGPT’s vulnerability to phishing attacks via prompt injection poses a significant security risk.

Also: BleepingComputer · The Register Sec · The Register Sec

AI-Powered Attacks Target Ukraine

The Hacker News · 2026-05-29 · corroborated · reporting

GREYVIBE’s use of AI-powered attacks against Ukraine highlights the growing threat of AI-assisted offense.

Also: The Register Sec

AI-Generated Code Not Accepted by SQLite

Simon Willison · 2026-05-27 · single-source · research

SQLite’s rejection of AI-generated code raises questions about the reliability and security of AI-produced software.

Claude Sandboxing Techniques

Simon Willison · 2026-05-30 · single-source · research

Anthropic’s Claude sandboxing techniques demonstrate a promising approach to containing AI agents.

AI-Assisted Bug Hunting Boosts Security

The Hacker News · 2026-05-23 · corroborated · reporting

The use of AI-powered bug hunting has significantly improved vulnerability discovery and security.

Also: The Register Sec

Copilot Cowork Leaks Files via Prompt Injection

Simon Willison · 2026-05-26 · single-source · research

The vulnerability of Copilot Cowork to prompt injection attacks highlights the need for improved security measures in AI-assisted development tools.

AI Security Incidents Affect Half of Organizations

The Register Sec · 2026-05-27 · corroborated · reporting

The prevalence of AI-related security incidents underscores the importance of addressing AI security risks.

Also: The Register Sec

Meta AI Bot Vulnerable to Password Reset Tricks

Krebs on Security · 2026-06-01 · single-source · reporting

The vulnerability of Meta’s AI-powered support bot to password reset tricks poses a significant security risk.

OpenAI Codex Tokens Stolen in Supply Chain Attack

The Hacker News · 2026-06-01 · single-source · reporting

The theft of OpenAI Codex tokens in a supply chain attack highlights the need for improved security measures in AI development.

Malicious npm Package Targets Claude AI User Files

The Hacker News · 2026-05-27 · single-source · reporting

The discovery of a malicious npm package targeting Claude AI user files underscores the importance of securing AI-related infrastructure.

AI Chatbots Spread Cryptojacking Malware

The Hacker News · 2026-05-27 · single-source · reporting

The use of AI chatbots to spread cryptojacking malware highlights the growing threat of AI-assisted attacks.

AI-Powered DDoS Attacks Become More Powerful

The Hacker News · 2026-05-26 · single-source · reporting

The increasing power of AI-powered DDoS attacks poses a significant threat to online security.

CERT-In Urges Faster Patching Due to AI-Assisted Attacks

The Hacker News · 2026-05-26 · corroborated · reporting

The need for faster patching in response to AI-assisted attacks highlights the importance of proactive security measures.

Also: The Register Sec

Megalodon Attacks GitHub CI/CD Workflows

The Hacker News · 2026-05-22 · single-source · reporting

The vulnerability of GitHub CI/CD workflows to Megalodon attacks underscores the need for improved security in AI-related development infrastructure.

Gemini Jailbreak Uses AI in Hacking Spree

The Register Sec · 2026-05-22 · single-source · reporting

The use of AI in the Gemini jailbreak highlights the growing threat of AI-assisted attacks.

Microsoft Opensources AI Safety Tools

The Register Sec · 2026-05-21 · single-source · reporting

Microsoft’s open-sourcing of AI safety tools demonstrates a commitment to improving AI security.

Undisclosed AI Sandbox Vulnerability Poses Risk

The Register Sec · 2026-05-20 · single-source · reporting

The existence of an undisclosed AI sandbox vulnerability highlights the need for improved security measures in AI development.

AI Speeds Up Vulnerability Discovery

Schneier · 2026-06-01 · single-source · analysis

The use of AI in vulnerability discovery has significant implications for security disclosure dynamics.

AI Aids macOS Exploit

Schneier · 2026-05-21 · single-source · analysis

The use of AI in exploiting macOS vulnerabilities highlights the growing threat of AI-assisted attacks.

Benchmarks Don’t Measure AI Security

Schneier · 2026-05-20 · single-source · analysis

The limitations of benchmarks in measuring AI security underscore the need for more comprehensive evaluation methods.

Nx Console Supply Chain Vulnerability

CISA Advisories · 2026-05-28 · single-source · advisory

The discovery of a supply chain vulnerability in Nx Console highlights the importance of securing AI-related infrastructure.

Claude Opus 4.8 Improves Honesty

Simon Willison · 2026-05-28 · single-source · research

The improvement of honesty in Claude Opus 4.8 demonstrates progress in AI development.

curl Team Overwhelmed by AI-Assisted Reports

Simon Willison · 2026-05-26 · single-source · research

The overwhelming of the curl team by AI-assisted reports highlights the need for improved security measures in AI-related development infrastructure.

Monty Sandbox Shows Promise

Simon Willison · 2026-05-22 · single-source · research

The Monty sandbox demonstrates a promising approach to containing AI agents.

AI Writes Cisco Reports with Errors

The Register Sec · 2026-05-22 · single-source · reporting

The use of AI in writing Cisco reports with errors highlights the limitations of AI-assisted development tools.

GlassWorm Malware Affects Dev Supply Chain

The Hacker News · 2026-05-27 · single-source · reporting

The discovery of GlassWorm malware affecting the dev supply chain underscores the importance of securing AI-related infrastructure.

AI Security Incidents Require Improved Measures

The Register Sec · 2026-05-27 · single-source · reporting

The prevalence of AI-related security incidents highlights the need for improved security measures.

AI-written emails

Simon Willison · 2026-05-26 · single-source · research

Security engineers should care about AI misuse

LLM agent used

The Hacker News · 2026-05-29 · single-source · reporting

AI-assisted offense expands attack scope