Close

Explore AI

Explore GPTs

Explore AI News

Explore AI Videos

Explore AI for Jobs

Latest AES AI Videos

Find the latest for AES company videos

Follow
... 1:4:21
Python + AI: Vector embeddings

Microsoft Reactor

1+ months ago 2K views

Vector embeddings are a critical topic in artificial intelligence, crucial for the representation of data inputs such as text, images, and audio into lists of floating point numbers for use in various applications. Different embedding models have been discussed, including Word2Vec, OpenAI's latest embedding models, and ESMERT. Understanding the various types of embeddings and their properties, including optimization techniques like quantization and dimensionality reduction, is essential for leveraging these technologies. This session emphasized how vector embeddings can be employed for similarity tracking and recommendation systems, underscoring their significance in modern AI applications.
03:21 - 03:24

Overview of embedding generation and practical demonstrations with GitHub.

Machine Learning Education
... 26:38
Hacking LLMs Demo and Tutorial (Explore AI Security Vulnerabilities)

David Bombal Tech

1+ months ago 8K views

AI vulnerabilities pose significant risks, intertwining traditional network security threats with unique issues inherent to AI systems. Ensuring data integrity and secure interactions with APIs is crucial, particularly as organizations integrate large language models (LLMs) and other AI technologies into their operations. The OWASP Foundation now addresses threats specific to LLMs, such as prompt injection and data poisoning. Resources like Cisco's free tutorial on AI security are essential for educating professionals about potential vulnerabilities and mitigation techniques. As AI continues to evolve rapidly, staying informed and proactive in security measures is vital for safeguarding organizations.
02:06 - 02:10

Rapid AI development increases risks and potential attack vectors.

Cisco Ethical AI frameworks Cybersecurity
... 37:34
Enterprise Linux Security Episode 110: AI Wars

Learn Linux TV

1+ months ago 2K views

AI is reshaping cybersecurity, with adversaries leveraging it to create sophisticated malware, evidenced by the emergence of the ransomware group 'Func.' This group has employed AI to amplify their attack capabilities, leading to higher victim counts. Concurrently, AI tools are being utilized by security researchers to combat these threats. The discussion reveals concerns about the accessibility of AI for malicious use, making cybercrime easier for lower-skilled actors. Another malware, 'Autocolor,' is highlighted, which employs evasion techniques and has a notable persistence mechanism, demonstrating the ongoing arms race between cybersecurity professionals and threat actors.
06:25 - 06:29

AI assists low-skilled hackers in quickly iterating on malware solutions.

Cybersecurity Cybersecurity
... 7:58
Don’t Push Sensitive Data To DeepSeek/OpenAI/Etc - Sanitize your Pcaps!!

Chris Greer

1+ months ago 2K views

This video explains how to effectively sanitize network traffic captures before sharing or uploading them to AI models for analysis. It emphasizes the importance of protecting sensitive data and highlights the use of a Windows tool called Trace Wrangler for this purpose. The process includes removing unwanted payloads, anonymizing IP addresses, and maintaining necessary information for analysis while ensuring that confidential data is not exposed. The speaker discusses various options available in Trace Wrangler to facilitate this sanitization process, ensuring compliance with security practices in network traffic sharing.
07:18 - 07:29

AI model analysis requires careful sanitization of shared traffic.

Deepseek Cybersecurity
... 12:46
DANGEROUS "EMOJI HACK": AI models susceptible to 'trojan horse' emojis...

Wes Roth

1+ months ago 28K views

A serious threat to AI security has been discovered involving emoji, particularly how certain seemingly simple emoji can encode significant amounts of hidden data through Unicode variation selectors. These structures can conceal instructions unknown to users, allowing for potential prompt injection attacks where malicious commands are smuggled into AI queries. The implications of this vulnerability may necessitate immediate action from developers to safeguard against such injections in AI models, as the techniques behind encoding and decoding information in this manner present new challenges in ensuring the integrity and security of language models.
02:40 - 02:50

Encoding data in Unicode characters reveals risks of information concealment.

Natural Language Processing (NLP) AI Trends
... 8:13
Ransomware Group Dismantled, AI Platform & Hipshipper Leak | cybernews.com

Cybernews

1+ months ago 5K views

Recent cyber news reveals significant data breaches involving the Israeli and American police forces, with Iranian hackers stealing sensitive information from Israeli authorities and a U.S. hacker group breaching Lexipol, exposing police training documents. Additionally, over 34 million messages from OmniGPT, a popular AI productivity platform, were exfiltrated, highlighting the risks of sharing sensitive data with AI assistants. The dismantling of the Eight-base ransomware group shows law enforcement action against cybercrime. Meanwhile, a summit discussed AI regulation without significant U.S. or U.K. participation, underscoring ongoing tensions in global AI governance.
03:10 - 03:57

OmniGPT's data breach exposed 34 million messages and sensitive user information.

Hipshipper Cybersecurity Cybersecurity
... 8:48
Playstation Outage, OpenAI Deep Research & DeepSeek | Weekly News

Cybernews

1+ months ago 5K views

Concerns over the use of the AI newscaster Joe reflect a growing consensus against its deployment. Recent cyber events highlight a hacker's claim to hack Trump Hotels, numerous issues with meme coins linked to high-profile individuals, and PlayStation Network outages that frustrated gamers. Additionally, a significant rise in Apple-targeted info-stealers was reported, alongside OpenAI's launch of a new product aimed at competing with DeepSeek. Issues with DeepSeek's vulnerabilities led to calls for regulatory action, including proposed bans by various governments due to data security concerns and its ties to Chinese AI systems.
OpenAI Cybersecurity Cybersecurity
... 8:48
Playstation Outage, OpenAI Deep Research & DeepSeek | Weekly News

Cybernews

1+ months ago 5K views

Concerns over the use of the AI newscaster Joe reflect a growing consensus against its deployment. Recent cyber events highlight a hacker's claim to hack Trump Hotels, numerous issues with meme coins linked to high-profile individuals, and PlayStation Network outages that frustrated gamers. Additionally, a significant rise in Apple-targeted info-stealers was reported, alongside OpenAI's launch of a new product aimed at competing with DeepSeek. Issues with DeepSeek's vulnerabilities led to calls for regulatory action, including proposed bans by various governments due to data security concerns and its ties to Chinese AI systems.
OpenAI Cybersecurity Cybersecurity