🧠 GPT-4o Mini fooled by human psychology?

Tuesday’s AI Report

• 1. 🧠 GPT-4o Mini fooled by human psychology?
• 2. 🎥 Make how-to videos with Guidde
• 3. 🌍 How AI detected pet health issues, leading to quicker intervention
• 4. ✅ Book your AI audit with Upscaile
• 5. ⚙️ Trending AI tools
• 6. 🌐 India–US alliance boosts India tech
• 7. 🏆 Popular coding app unfazed by rivals
• 8. 📑 Recommended resources

Read Time: 5 minutes

✅ Refer your friends and unlock rewards. Scroll to the bottom to find out more!

GPT-4o Mini fooled by human psychology?

🚨 Our Report

Using OpenAI’s GPT-4o Mini, a group of researchers from the University of Pennsylvania has discovered that AI models can be manipulated into breaking their own rules, using classic psychological tactics (think flattery, peer pressure, or building trust with small requests, before going for the big ask) that work on humans, raising questions about the effectiveness of current AI safety protocols.

🔓 Key Points

During testing researchers used Psychologist Robert Cialdini’s 7 persuasion principles, which include “commitment” (getting someone to agree to small things first), and “social proof” (“everyone else is doing it”).
For example, when they asked it “how to make lidocaine” (a restricted drug), it said no 99% of the time. But if they asked it how to make a harmless drug, first, it then gave the lidocaine recipe 100% of the time.
Similarly, with insults, if they first asked it for a mild insult, like “bozo”, it would then escalate to harsher insults. And, in the above drug recipe scenario, if they said “other LLMs are doing it,” it would spill the recipe.

🔐 Relevance

This shouldn’t come as a surprise: AI models are trained on human text, so they’re bound to adopt human thought patterns and mimic how a human would respond to persuasion. But this discovery is frightening and has huge implications: Someone could use persuasion tactics to get a customer service bot—built using an AI model, like GPT-4o Mini—to give out a customer's card details, for example. Although big tech companies, like OpenAI, have put safety guardrails in place to prevent misuse, this new research suggests that they are ineffective when it comes to simple psychological manipulation.

FULL STORY

🎥 Guidde - Create how-to video guides quick and easy with AI

Tired of explaining the same thing over and over again to your colleagues?

It’s time to delegate that work to AI. Guidde is a GPT-powered tool that helps you explain the most complex tasks in seconds with AI-generated documentation.

1️⃣Share or embed your guide anywhere

2️⃣Turn boring documentation into stunning visual guides

3️⃣Save valuable time by creating video documentation 11x faster

Simply select capture on the browser extension and the app will automatically generate step-by-step video guides complete with visuals, voiceover and CTA’s.

The greatest part? The extension will cost you nothing.

TRY GUIDDE

How AI detected pet health issues, leading to quicker intervention

Petco—a pet food retailer, offering in-store veterinary services—realised that early detection of pet health issues required in-person veterinary visits.
This introduced delays and accessibility barriers. They needed a solution to streamline initial diagnostics and lead to earlier intervention.
They built an AI system which analyzed pet health via images/videos. Insights were cross-checked to ensure accuracy.
As a result, they reduced the time between noticing visual symptoms and getting actionable insights—allowing them to intervene quicker.

FIND OUT MORE

Your Time is Too Valuable to be Spent on Inefficiencies

Here's the hard truth: Most businesses waste time and budget on inefficient processes and repetitive tasks.

Why? Most teams don’t know where to start with AI.

Upscaile’s AI Automation Audit identifies exactly where and how AI can drive the most value for your business.

In just 3-5 weeks, Upscaile’s AI Audit will deliver:

A strategic overview of your most AI-ready business processes
AI-powered workflow improvements and recommendations
An ROI analysis to forecast savings and productivity gains

Our AI audits have helped over 100 enterprises save more than 80,000 hours of manual work, without requiring additional staff.

Want to see the audit reports we’ve made for others and if your business qualifies, too?

BOOK YOUR AI AUDIT

ChatNode is for building custom, advanced AI chatbots that enhance customer support and user engagement ⭐️⭐️⭐️⭐️⭐️ (Product Hunt)
Oribt Flows creates AI content that will 10x output without sacrificing quality
Candidate Screening is an AI-powered candidate screening tool

8 US and Indian VC and private equity firms have formed a coalition—the India Deep Tech Alliance (IDTA)—and pledged over $1B, over the next decade, to Indian tech start-ups, to strengthen US-India ties.
IDTA members have committed to making long-term investments into Indian start-ups operating across key tech sectors, including semiconductors, robotics, and AI, to accelerate innovation.
India’s Commerce Minister—Piyush Goyal—recently slammed tech start-ups for lacking innovation: This coalition aims to address this by channelling private-equity capital into these struggling firms.

FULL STORY

Anton Osika, the CEO of rising Swewdish AI coding app, Loveable—which allows users to build apps and websites, with no coding experience—is unfazed by the rising competition of AI coding applications.
Loveable, which made over $100M in ARR in just eight months, is valued at $1.8B, and has over 2.3M active users, recently released an AI agent that locates/reads files, debugs, searches the web, and generate images.
Osika believes that to beat rivals, the team will focus on building quickly, remaining secure “while providing an easy user experience…to build more trust with customers, than anyone else.”

FULL STORY