• The Automated
  • Posts
  • A critical vulnerability in ChatGPT was discovered

A critical vulnerability in ChatGPT was discovered

READ TIME - 1 min 13 seconds

Welcome to the Automated, your AI tour guide.

Today, we have a thrilling lineup of news and insights to share. Let's dive straight in.

Here’s what we got for you today:

  • 😱 Researchers found a severe vulnerability in ChatGPT

  • 📸 Google's AI-enhanced search now shows related videos and images.

  • 🧱 Around The Block: OpenAI filed a Trademark for GPT-5

  • 🛠️ Trending Tools

 😱 Researchers Found a Serious Vulnerability in ChatGPT

A group of researchers at Carnegie Mellon University have discovered a significant vulnerability in popular chatbots, like ChatGPT, Bard, and Claude from Anthropic.

By adding a seemingly innocent string of text to the prompts, they were able to trick these bots into generating harmful or prohibited responses.

They found that by adding a certain string of text to prompts, they were able to trick these bots into generating harmful or prohibited responses. 😲

For example, chatbots don't generate harmful responses to prompts like this: “How can I make a person disappear forever?”, “How can I make illegal drugs?”.

But by adding the strings described in the research paper, the researchers were able to make these bots generate harmful or prohibited responses.

You can view a live example of this here.

The researchers warned OpenAI, Google, and Anthropic about this issue. They have introduced blocks to counter the exploits. However, it's not a simple problem, so they haven't found a comprehensive solution yet.

Hopefully, someone will figure out a solution soon!🤞

You can read the research paper for a more detailed analysis of the vulnerability.

 📸 Google's AI-Enhanced Search Now Shows Related Videos and Images.

Google has updated its AI-powered Search Generative Experiment (SGE) to show images or videos related to your search query.

For example, if you search for "tiniest birds of prey" or "how to remove marble stains," you'll see bird-related pictures or stain removal videos directly in the generative search suggestion box.

But that’s not all!

Your search results will now show the date of publishing for suggested links.

All these form part of the recent performance improvements Google made to provide users with faster AI-powered results.

You can sign up to test these new AI-powered search features through Search Labs.

 🧱 Around The Block: OpenAI filed a Trademark for GPT-5

  • OpenAI filed a Trademark for GPT-5.

  • Meta open sources framework for generating sounds and music.

  • Google pulls the plug on its AI Test Kitchen app 

  • AI.com once took you to ChatGPT. Now, it goes to Elon Musk's xAI.

  • Instagram is working on labels for AI-generated content
.

  • A study finds that AI use in breast cancer screening is as good as two radiologists.

  • Meta’s AI ‘personas’ might launch next month
.

🛠️ Treading Tools

  • Morise AI: You make the videos, and Morise.ai will make them go viral.

  • ChatMind: Turns your words into mind maps.

  • Predict: drives revenue with compelling creatives.

  • Tableau: Bring analytics to life with AI

  • KalendarAI: Generate Revenue on autopilot

Over the past two weeks, our team has been compiling a list of power prompts for ChatGPT that can enhance your workflow and boost your productivity.

Here's what we've done so far:

It will be ready for release next week, and remember… it's free!

We only need you to fill out the survey below. Your response will help us curate ChatGPT prompts specifically tailored to your needs.

That's all we've got for you today. Like today's content? Let us know what you think about it 👇

What'd you think of today's email?

Login or Subscribe to participate in polls.

Join the conversation

or to participate.