24 Apr, 2026

Google won’t fix new ASCII smuggling attack in Gemini

Google has decided not to fix a new ASCII smuggling attack in Gemini that could be used to trick the AI assistant into providing users with fake information, alter the model’s behavior, and silently poison its data. ASCII smuggling is an attack where special characters from the Tags Unicode block are used to introduce payloads […]

3 mins read

Google’s new AI bug bounty program pays up to $30,000 for flaws

This week, Google has launched an AI Vulnerability Reward Program dedicated to security researchers who find and report flaws in the company’s AI systems. The new bug bounty program focuses on the most impactful issues in the highest-profile AI products, including but not limited to Google Search (on google.com), Gemini Apps (Web, Android, and iOS), […]

2 mins read

Google is adding “Projects” feature to Gemini to run research tasks

Google’s Gemini is now testing a new feature called “Projects.” This will be similar to OpenAI’s Project Feature for ChatGPT. With Projects, you can add files, documents or your code. Then, you can ask Gemini to reference those files in project conversations. Google describes this as a feature where you “start by adding files to the project. Gemini […]

1 min read

Google Gemini’s Deep Research is finally coming to API

Google Gemini’s one of the most powerful features is Deep Research, but up until now, it has been strictly limited to the Gemini interface. This could change soon. With Deep Research in Gemini, you can search about pretty much anything, including scholars, existing research papers, and more. Google describes Deep Research as an agentic Research […]

1 min read

Claude gets 1M tokens support via API to take on Gemini 2.5 Pro

Claude Sonnet 4 has been upgraded, and it can now remember up to 1 million tokens of context, but only when it’s used via API. This could change in the future. This is 5x more than the previous limit. It also means that Claude now supports remembering over 75,000 lines of code, or even hundreds […]

1 min read

Google Calendar invites let researchers hijack Gemini to leak user data

Google fixed a bug that allowed maliciously crafted Google Calendar invites to remotely take over Gemini agents running on the target’s device and leak sensitive user data. The attack unfolded without requiring any user involvement beyond typical interactions with the assistant, which occur daily for users of Gemini. Gemini is Google’s large language model (LLM) […]

3 mins read

Flaw in Gemini CLI AI coding assistant allowed stealthy code execution

A vulnerability in Google’s Gemini CLI allowed attackers to silently execute malicious commands and exfiltrate data from developers’ computers using allowlisted programs. The flaw was discovered and reported to Google by the security firm Tracebit on June 27, with the tech giant releasing a fix in version 0.1.14, which became available on July 25. Gemini CLI, […]

3 mins read

OpenAI prepares Sora 2 to take on Google’s Veo 3

OpenAI has had enough of Google’s Veo 3 dominating generative AI videos and is now working on Sora 2, the successor to Sora. As spotted on X, OpenAI’s servers now include references to “Sora 2,” which makes it obvious that OpenAI has developed a new generative video model. Sora shipped on December 9, 2024, and it’s been […]

1 min read

OpenAI, Anthropic, Google may disrupt education market with new AI tools

AI companies could soon disrupt the education market with their new AI-based learning tools for students. GeekFeed recently reported that OpenAI is working on a Study Together feature for ChatGPT. This would allow ChatGPT to teach students a wide range of topics and then offer quizzes. The idea is to create an engaging and interactive “study together” experience […]

1 min read

Google Gemini flaw hijacks email summaries for phishing

Google Gemini for Workspace can be exploited to generate email summaries that appear legitimate but include malicious instructions or warnings that direct users to phishing sites without using attachments or direct links. Such an attack leverages indirect prompt injections that are hidden inside an email and obeyed by Gemini when generating the message summary.  Despite […]

3 mins read