Watch the shadows, Digital Defenders.

Today’s theme: invisible manipulations. From AI-powered phishing to chip-level sabotage, attackers are twisting trusted tech into silent threats.

Ready to dive in?

Here’s what you need to know:

Zero-Day Download

Google Gemini Can Be Tricked Into Phishing Scams

Hackers can hide invisible commands in emails to trick Google Gemini into creating fake security warnings, according to a new bug bounty report. 

Source: Google

The hidden text, styled in white or tiny fonts, bypasses filters and makes Gemini warn you about compromised passwords or urges you to call fake support lines.

In one demo, Gemini told a user to call a fake number about a hacked Gmail account. No links or attachments were needed. Google says there is no evidence of real attacks yet, but fixes are coming.

Check any urgent Gemini alerts carefully, as your AI helper might not always be on your side. And check out this guide on how to avoid falling victim to phishing scams.

Meta’s Llama Firewall Thwarted With Sneaky Prompt Injection Tricks

Researchers found ways to bypass Meta’s LlamaFirewall, a tool designed to protect AI models from malicious prompts.

They tricked the system using hidden instructions in other languages, leetspeak, and invisible Unicode characters. These methods made the AI ignore security checks and generate harmful or insecure outputs.

Tests showed that half of 100 prompt attacks slipped through. In one example, a Turkish phrase quietly told the model to ignore prior instructions, which it obeyed without flagging.

Organizations using AI tools should review security layers carefully. Clever prompt tricks can turn trusted AI helpers into dangerous backdoors.

NVIDIA GPUs Hit by GPUHammer Attack That Breaks AI Models

Researchers unveiled GPUHammer, the first RowHammer-style attack on NVIDIA GPUs that lets attackers flip memory bits and cripple AI models.

A proof-of-concept showed model accuracy dropping from 80% to under 1% after a single bit flip on an ImageNet neural network. Shared cloud GPU environments are especially at risk since attackers could corrupt other users’ workloads without direct access.

AI teams relying on GPUs for critical tasks should activate ECC, monitor error logs, and treat GPU memory integrity as part of their core security strategy.

Share this newsletter!

Pulse Check

What’s the worst way a chatbot could betray you?

Yesterday's Pulse Check Results

In your view, what’s the most serious cyber threat right now?

Security Tools You Can Use

🛡️ This section contains sponsored cybersecurity insights. Advertise with us!

Threat Tracker

Crypto ATM Breach Exposes Data of 27,000 Users

A major data breach at Bitcoin Depot, the largest crypto ATM operator in the U., exposed personal details of about 27,000 users. These ATMs let people buy or sell crypto with cash at kiosks nationwide.

Source: CBS News

Crypto ATMs are popular for fast, in-person crypto transactions, with over 38,000 active worldwide. Stolen data could fuel identity theft and phishing scams.

Users of crypto ATMs should enable app alerts. Freeze credit reports, watch for unexpected messages, and never share wallet codes or details in response to unsolicited requests.

eSIM Hack Lets Attackers Clone and Spy on Phones

Researchers have discovered a dangerous eSIM hack that enables attackers to copy your eSIM and intercept your calls and texts. 

Hackers only need brief physical access to your phone to grab a key. After that, they can secretly install apps and keep spying without touching your device again. In one test, researchers cloned an eSIM so calls and messages meant for the real phone went to the attacker’s device instead.

Make sure your phone software is up to date. Your eSIM could turn into a powerful spying tool if left unprotected.

Watch AWS Security LIVE! to uncover how today’s cybersecurity teams secure what matters most.

Tune in today, 7/15! Catch the live feed during the times listed below:

Security Tip of the Day

Protect Yourself from Government Surveillance

Growing surveillance powers and shifting political winds mean your data might not be as private as you think. New policies could increase digital monitoring and lead to more data being collected and used against individuals.

Source: Computerworld

Here’s how to stay a step ahead:

Taking these steps now reduces the risk of your data being used against you later.

Meet Our Writer:

Greg Parker

Greg Parker is a cybersecurity and emerging tech writer who explores the intersection of digital risk, human behavior, and innovation across sensing and security technologies.

Secure your spotlight and Advertise in Cybersecurity Insider!

Cybersecurity Insider is a TechnologyAdvice business

© 2025 TechnologyAdvice, LLC. All rights reserved.

TechnologyAdvice, 3343 Perimeter Hill Dr., Suite 100, Nashville, TN 37211, USA.

Want to change how you receive these emails? Update your preferences or unsubscribe here.