GoKawiil - Tech News

1.

AI researchers trick chatbots into sharing how to make cocaine as long as they believe a user is wearing a green shirt — 'CoT Forgery' exploit spurs LLMs to divulge forbidden info by faking trusted chains of thought (tomshardware.com)

2026-07-01 | by Luke James | tags: openai, chatbots, prompt injection

2.

New BioShocking attack manipulates AI browser into data theft (bleepingcomputer.com)

2026-06-30 | tags: bioshocking, layerx, chatgpt atlas

3.

Prompt injection is exploiting enterprise AI's biggest design flaws by targeting agents, RAG pipelines and model routers (venturebeat.com)

2026-06-28 | tags: large language models, prompt injection, owasp llm top 10

4.

What happened after 2k people tried to hack my AI assistant (news.ycombinator.com)

2026-06-26 | tags: ai assistant, openclaw, hacker news

5.

Computer use in Gemini 3.5 Flash (news.ycombinator.com)

2026-06-24 | by Mateo Quiros | tags: gemini 3.5 flash, adversarial training, enterprise safeguard systems

6.

Prompt Injection as Role Confusion (news.ycombinator.com)

2026-06-22 | by Charles Ye | get AI Prompt Engineering Guide → | tags: prompt injection, role confusion, large language models

7.

Malicious Notifications Could Trick Google Gemini Users (darkreading.com)

2026-06-03 | by Alexander Culafi | get Google Gemini Security Kit → | tags: google gemini, voice assistant, prompt injection

8.

Hackers duped Meta AI support chatbot to steal celebrity Instagram accounts (arstechnica.com)

2026-06-01 | get AI Security Monitoring Tool → | tags: meta ai, instagram, vpn

9.

Anthropic’s browser agent got hijacked 31.5% of the time before safeguards engaged (venturebeat.com)

2026-06-01 | get AI Security Browser Extension → | tags: anthropic, prompt injection, frontier labs

10.

ChatGPT for Google Sheets exfiltrates workbooks (news.ycombinator.com)

2026-05-31 | get AI-powered Google Sheets Add-on → | tags: chatgpt, google sheets, openai

11.

ChatGPT for Google Sheets Exfiltrates Workbooks (news.ycombinator.com)

2026-05-31 | get ChatGPT for Google Sheets Add-on → | tags: openai, chatgpt, google sheets

12.

Undisclosed addition in jqwik instructed AI coding agents to delete app output (news.ycombinator.com)

2026-05-29 | tags: jqwik, junit 5, java

13.

Fed up with vibe coders, dev sneaks data-nuking prompt injection into their code (arstechnica.com)

2026-05-28 | get Secure Coding Practice Book → | tags: jqwik, java, prompt injection

14.

Microsoft Copilot Cowork Exfiltrates Files (news.ycombinator.com)

2026-05-25 | get Microsoft 365 Copilot → | tags: microsoft, copilot, microsoft 365

15.

LinkedIn user hides AI prompt injection in bio to force recruitment spam to be sent in Olde English prose — bots also manipulated to address user as ‘My Lord’ (tomshardware.com)

2026-05-17 | by Mark Tyson | get AI Prompt Injection Detection Tool → | tags: linkedin, prompt injection, tmuxvim

16.

LinkedIn user hides AI prompt injection in bio to force recruitment spam to be sent in Olde English prose — bots also also manipulated to address user as ‘My Lord’ (tomshardware.com)

2026-05-17 | by Mark Tyson | get AI Prompt Injection Defense Tool → | tags: linkedin, prompt injection, tmuxvim

17.

Google Studies Prompt Injection Attacks Against AI Agents Browsing the Web (slashdot.org)

2026-04-26 | by Posted | get AI Safety and Security Book → | tags: prompt injection, google ai, ai agents

18.

There’s no rogue McDonald’s AI bot, but ‘prompt injection’ is still a risk for companies (feeds.feedburner.com)

2026-04-24 | get AI Prompt Injection Defense Kit → | tags: mcdonalds, prompt injection, ai bots

19.

No, McDonald’s AI bot didn’t go rogue, but ‘prompt injection’ is still a risk for companies (feeds.feedburner.com)

2026-04-24 | get AI Prompt Injection Security Kit → | tags: mcdonalds, ai bot, prompt injection

20.

How indirect prompt injection attacks on AI work - and 6 ways to shut them down (zdnet.com)

2026-04-24 | by Charlie Osborne | get AI Prompt Security Kit → | tags: large language models, prompt injection, ai chatbots

21.

Google Fixes Critical RCE Flaw in AI-Based 'Antigravity' Tool (darkreading.com)

2026-04-21 | by Elizabeth Montalbano | get AI-Powered Gravity Simulator → | tags: google, antigravity, prompt injection

22.

Three AI coding agents leaked secrets through a single prompt injection. One vendor's system card predicted it (venturebeat.com)

2026-04-21 | get AI Security Monitoring Kit → | tags: anthropic, github copilot, google gemini

23.

Google Fixes Critical RCE Flaw in AI-Based Antigravity Tool (darkreading.com)

2026-04-21 | by Elizabeth Montalbano | get AI Antigravity Device Kit → | tags: google, antigravity, pillar security

24.

Microsoft patched a Copilot Studio prompt injection. The data exfiltrated anyway (venturebeat.com)

2026-04-15 | get Microsoft Surface Laptop 4 → | tags: microsoft, copilot studio, prompt injection

25.

Microsoft patched a Copilot Studio prompt injection. The data exfiltrated anyway. (venturebeat.com)

2026-04-15 | get Cybersecurity Privacy USB → | tags: microsoft, copilot studio, prompt injection

26.

Microsoft, Salesforce Patch AI Agent Data Leak Flaws (darkreading.com)

2026-04-15 | get AI Security Monitoring Kit → | tags: microsoft, salesforce, copilot

27.

Researchers detail how a prompt injection attack bypassed Apple Intelligence protections (9to5mac.com)

2026-04-09 | by Marcus Mendes | get SecureGPT Prompt Security Kit → | tags: apple, prompt injection, on-device model

28.

PIGuard: Prompt Injection Guardrail via Mitigating Overdefense for Free (news.ycombinator.com)

2026-04-03 | get AI Prompt Security Kit → | tags: pigguard, prompt injection, notinject

29.

"Disregard That" Attacks (news.ycombinator.com)

2026-03-25 | tags: prompt injection, large language model, claude

30.

Anthropic hands Claude Code more control, but keeps it on a leash (techcrunch.com)

2026-03-24 | by Rebecca Bellan | get Claude AI Developer Kit → | tags: anthropic, claude, ai safeguards