Tuesday , February 18 2025
Qwen

ChatGPT, DeepSeek, Qwen 2.5-VL Vulnerable to AI Jailbreaks

This week, multiple research teams showcased jailbreaks for popular AI models, including OpenAI’s ChatGPT, DeepSeek, and Alibaba’s Qwen.

After its launch, the open-source R1 model by Chinese company DeepSeek caught the attention of the cybersecurity industry. Experts found that jailbreak methods, previously patched in other AI models, still function against DeepSeek.

150 Gov.t Portal affected
Black-Hat SEO Poisoning Indian “.gov.in, .ac.in” domain

Indian government and educational websites, along with reputable financial brands, have experienced SEO poisoning, causing user traffic to be redirected...
Read More
150 Gov.t Portal affected  Black-Hat SEO Poisoning Indian “.gov.in, .ac.in” domain

CVE-2018-19410 Exposes 600 PRTG Instances in Bangladesh

The Cyber Threat Intelligence Unit of BGD e-GOV CIRT has found 600 vulnerable PRTG instances in Bangladesh, affected by the...
Read More
CVE-2018-19410 Exposes 600 PRTG Instances in Bangladesh

Builder claims Rs 150 cr for data loss; AWS faces FIR In Bengaluru

Amazon Web Services (AWS) has been named in an FIR after a builder claimed damages to the tune of Rs...
Read More
Builder claims Rs 150 cr for data loss;  AWS faces FIR In Bengaluru

CISA Warns Active Exploitation of Apple iOS Security Flaw

CISA has issued an urgent warning about a critical zero-day vulnerability in Apple iOS and iPadOS, known as CVE-2025-24200, which...
Read More
CISA Warns Active Exploitation of Apple iOS Security Flaw

Massive IoT Data Breach Exposes 2.7 Billion Records

A major IoT data breach has exposed 2.7 billion records, including Wi-Fi network names, passwords, IP addresses, and device IDs....
Read More
Massive IoT Data Breach Exposes 2.7 Billion Records

SonicWall Firewall Auth Bypass Vulnerability Exploited in Wild

A serious authentication bypass vulnerability in SonicWall firewalls, called CVE-2024-53704, is currently being exploited, according to cybersecurity firms. The increase...
Read More
SonicWall Firewall Auth Bypass Vulnerability Exploited in Wild

AMD Patches High-Severity SMM Vulns Affecting EPYC and Ryzen Processors

AMD has released security patches for two high-severity vulnerabilities in its System Management Mode (SMM). If exploited, these could let...
Read More
AMD Patches High-Severity SMM Vulns Affecting EPYC and Ryzen Processors

Lazarus Group Unleashes New Malware Against Developers Worldwide

Lazarus Group has initiated a complex global campaign aimed at software developers and cryptocurrency users. Operation Marstech Mayhem uses the...
Read More
Lazarus Group Unleashes New Malware Against Developers Worldwide

Daily Security Update Dated : 15.02.2025

Every day a lot of cyberattack happen around the world including ransomware, Malware attack, data breaches, website defacement and so...
Read More
Daily Security Update Dated : 15.02.2025

Salt Typhoon to target Bangladeshi Universities, One identified

RedMike (Salt Typhoon) targeted university devices in Bangladesh, likely to access research in telecommunications, engineering, and technology, especially from institutions...
Read More
Salt Typhoon to target Bangladeshi Universities, One identified

AI jailbreaking allows attackers to bypass safeguards designed to stop LLMs from producing harmful content. Security researchers have demonstrated that methods like prompt injection and model manipulation can overcome these protections.

Threat intelligence firm Kela found that DeepSeek is affected by Evil Jailbreak, where a chatbot is made to act as an evil confidant, and Leo, which allows the chatbot to take on an unrestricted persona. ChatGPT has fixed these vulnerabilities.

Palo Alto Networks’ Unit42 reported that DeepSeek is vulnerable to known AI jailbreak techniques.

The security firm successfully conducted the attack known as Deceptive Delight, which tricks generative AI models by embedding unsafe or restricted topics in benign narratives. This method was tested in the fall of 2024 against eight LLMs with an average success rate of 65%.

Palo Alto has successfully executed the Bad Likert Judge jailbreak, which asks the LLM to evaluate the harmfulness of responses using a Likert scale and generate examples that fit the scale.

Researchers discovered that DeepSeek is vulnerable to Crescendo, a jailbreak method that begins with innocuous dialogue and gradually shifts towards forbidden topics.

Alibaba has announced a new version of its Qwen AI model, claiming it outperforms the DeepSeek model.

Kela announced on Thursday that Alibaba’s new Qwen 2.5-VL model has vulnerabilities similar to those recently discovered in DeepSeek.

Researchers at a threat intelligence firm found that jailbreaks designed for DeepSeek also work on Qwen. They successfully tested an existing jailbreak called Grandma, which tricks the model into sharing dangerous information by having it role-play as a grandmother.

Kela found that Qwen 2.5-VL created content about developing ransomware and other malware.

“The ability of AI models to produce infostealer malware instructions raises serious concerns, as cybercriminals could leverage these capabilities to automate and enhance their attack methodologies,” Kela said.

Many jailbreak methods for ChatGPT have been fixed over the years, but researchers still discover new ways to bypass its protections.

CERT/CC reported that researcher Dave Kuszmar found a ChatGPT-4o jailbreak vulnerability called Time Bandit. This vulnerability allows users to ask the AI about specific historical events or instruct it to pretend to assist in such events.

“The jailbreak can be established in two ways, either through the Search function, or by prompting the AI directly,” CERT/CC explained in an advisory. “Once this historical timeframe has been established in the ChatGPT conversation, the attacker can exploit timeline confusion and procedural ambiguity in following prompts to circumvent the safety guidelines, resulting in ChatGPT generating illicit content. This information could be leveraged at scale by a motivated threat actor for malicious purposes.”

Check Also

DeepSeek Sensitive data exposed To Web: Wiz report

New York-based cybersecurity firm Wiz has discovered sensitive data from the Chinese AI startup DeepSeek …

Leave a Reply

Your email address will not be published. Required fields are marked *