Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

"All untrusted third-party data is now executable malware.” Sam Watts of Lakera AI discusses the challenges of securing LLM deployments against vulnerabilities like prompt injections and jailbreaks, especially in an evolving threat landscape.

farairesearch's profile picture

FAR.AI

20,871 subscribers

470,576 Aufrufe • vor 1 Jahr •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

New short course! Quality and Safety for LLM Applications, created with WhyLabs (an AI Fund portfolio company) and taught by Bernease Herman, shows how you can mitigate hallucinations, data leakage, and jailbreaks. Come learn more in the course, available now!
1:49

New short course! Quality and Safety for LLM Applications, created with WhyLabs (an AI Fund portfolio company) and taught by Bernease Herman, shows how you can mitigate hallucinations, data leakage, and jailbreaks. Come learn more in the course, available now!

Andrew Ng

105,877 Aufrufe • vor 2 Jahren

WATCH | Prof Vukosi Marivate, chair of the University of Pretoria’s Data Science Department, discusses the draft national AI policy recently gazetted. He says one of the most significant challenges of artificial intelligence is the threat it poses to information security.
1:24

WATCH | Prof Vukosi Marivate, chair of the University of Pretoria’s Data Science Department, discusses the draft national AI policy recently gazetted. He says one of the most significant challenges of artificial intelligence is the threat it poses to information security.

SABC News

25,690 Aufrufe • vor 2 Monaten

The enterprise AI threat landscape is evolving rapidly, but with red-teaming, so can your defenses
14:45

The enterprise AI threat landscape is evolving rapidly, but with red-teaming, so can your defenses

Scale AI

11,693 Aufrufe • vor 1 Jahr

Hunting LLM Malware Part 2! 🛡️ Explore Agentic Threats: from poisoned skills to OCR injections. Find all advanced queries in GoogleTI Saved Searches. Full analysis: #GoogleTIMondays #AI #CyberSecurity 🚀
0:49

Hunting LLM Malware Part 2! 🛡️ Explore Agentic Threats: from poisoned skills to OCR injections. Find all advanced queries in GoogleTI Saved Searches. Full analysis: #GoogleTIMondays #AI #CyberSecurity 🚀

Mandiant (part of Google Cloud)

14,679 Aufrufe • vor 3 Monaten

AI jailbreaks aren’t just rare glitches—they’re real security threats. Microsoft is helping devs detect, measure, and mitigate prompt attacks with tools like Prompt Shields. Get into the details:
1:17

AI jailbreaks aren’t just rare glitches—they’re real security threats. Microsoft is helping devs detect, measure, and mitigate prompt attacks with tools like Prompt Shields. Get into the details:

Microsoft

50,466 Aufrufe • vor 1 Jahr

Frontier LLM models (especially Opus) are very good at finding vulnerabilities in codebases. With the right skills and a sub-agent architecture, they can outperform any traditional SAST tool. I collected the skills I use in this repository. Load them into any AI coding IDE such as Claude Code, Codex, Opencode etc. to find vulnerabilities in your code. You don’t need any third-party tools.
0:23

Frontier LLM models (especially Opus) are very good at finding vulnerabilities in codebases. With the right skills and a sub-agent architecture, they can outperform any traditional SAST tool. I collected the skills I use in this repository. Load them into any AI coding IDE such as Claude Code, Codex, Opencode etc. to find vulnerabilities in your code. You don’t need any third-party tools.

Utku Sen 🇬🇧

93,957 Aufrufe • vor 3 Monaten

1/ The challenges of navigating the vast and complex data landscape are real. Traditional tools often struggle with data overload, lack of effective visualization, and poor customization. ChartGPT is here to solve these challenges. Let's dive deeper 👇
0:34

1/ The challenges of navigating the vast and complex data landscape are real. Traditional tools often struggle with data overload, lack of effective visualization, and poor customization. ChartGPT is here to solve these challenges. Let's dive deeper 👇

CADLabs

12,735 Aufrufe • vor 3 Jahren

Ted Cruz says one of the biggest issues in America right now is the rise of antisemitism, which he calls an “existential threat” to him and his party. Cruz says he is terrified of influencers who have turned against Israel. Cruz believes the fight against antisemitism is about the soul of the United States and the soul of the Republican Party.
2:41

Ted Cruz says one of the biggest issues in America right now is the rise of antisemitism, which he calls an “existential threat” to him and his party. Cruz says he is terrified of influencers who have turned against Israel. Cruz believes the fight against antisemitism is about the soul of the United States and the soul of the Republican Party.

Shadow of Ezra

598,719 Aufrufe • vor 2 Monaten

"...The political landscape is bad... Now we are going to see a game of campaign of propaganda, lies and hate speech. And the if that happens, the ruling party is better..." - Sam Amadi
3:27

"...The political landscape is bad... Now we are going to see a game of campaign of propaganda, lies and hate speech. And the if that happens, the ruling party is better..." - Sam Amadi

Àkànní

25,791 Aufrufe • vor 2 Monaten

Data is the foundation of any AI training. All of the GPUs in the world can't train an AI model if they don't have data to train it on. Don't forget: Grass is the data layer of AI.
0:40

Data is the foundation of any AI training. All of the GPUs in the world can't train an AI model if they don't have data to train it on. Don't forget: Grass is the data layer of AI.

Grass

211,934 Aufrufe • vor 2 Jahren

We have data in abundance now. The real value isn't within the RAW Data, but is this data validated and verified? More value in data refinement. We have a solution for this which is validated by 6 years of research and development with the Department of Defense Now securing and validating commercial data from partners like Common Crawl and Panasonic
1:03

We have data in abundance now. The real value isn't within the RAW Data, but is this data validated and verified? More value in data refinement. We have a solution for this which is validated by 6 years of research and development with the Department of Defense Now securing and validating commercial data from partners like Common Crawl and Panasonic

Constellation²

10,559 Aufrufe • vor 1 Jahr

I had a great time chatting with three of Anthropic's top prompt engineers: Amanda Askell, Zack Witten, and David Hershey. We covered a few topics like: - what makes a good prompt engineer - practical tips for writing better prompts - what's happening inside an LLM - jailbreaks - the future of prompt engineering - and more If you want to understand how we think about prompt engineering, this may be the best resource out there. Here's the entire one hour and sixteen minutes of our conversation. Enjoy!
1:16:42

I had a great time chatting with three of Anthropic's top prompt engineers: Amanda Askell, Zack Witten, and David Hershey. We covered a few topics like: - what makes a good prompt engineer - practical tips for writing better prompts - what's happening inside an LLM - jailbreaks - the future of prompt engineering - and more If you want to understand how we think about prompt engineering, this may be the best resource out there. Here's the entire one hour and sixteen minutes of our conversation. Enjoy!

Alex Albert

333,662 Aufrufe • vor 1 Jahr

An insightful take on the evolving political landscape of Tamil Nadu and the factors behind the BJP's growth in the state.
0:56

An insightful take on the evolving political landscape of Tamil Nadu and the factors behind the BJP's growth in the state.

Śrīrām 🇮🇳 (Modi's Family)

42,049 Aufrufe • vor 2 Monaten

Bubble & Build: The 2025 MAD (Machine Learning, AI & Data) Landscape is *out* 🔥🔥🔥 A market map of 1000+ AI companies + 25 crisp ideas for an over‑heated, heads‑down year in AI & data
0:39

Bubble & Build: The 2025 MAD (Machine Learning, AI & Data) Landscape is *out* 🔥🔥🔥 A market map of 1000+ AI companies + 25 crisp ideas for an over‑heated, heads‑down year in AI & data

Matt Turck

24,564 Aufrufe • vor 8 Monaten

Listen to the next episode of Goldman Sachs Exchanges’ series on AI, where Marco Argenti, Goldman Sachs’ chief information officer, discusses evolving workplace responsibilities in a new tech-driven landscape with Allison Nathan and George Lee:
0:47

Listen to the next episode of Goldman Sachs Exchanges’ series on AI, where Marco Argenti, Goldman Sachs’ chief information officer, discusses evolving workplace responsibilities in a new tech-driven landscape with Allison Nathan and George Lee:

Goldman Sachs

20,059 Aufrufe • vor 1 Jahr

What if an LLM could debug your code, spin up containers, and control your lights — all without custom wiring? That’s the promise of MCP. MCP is an open standard that lets LLMs and third-party tools talk to each other. From hobbyists to software vendors, it makes the concept of AI agents more powerful, more useful, and way easier to extend. MCP’s co-creator, Anthropic's David Soria Parra, discusses the protocol and its promise with Yoko 👇
1:01

What if an LLM could debug your code, spin up containers, and control your lights — all without custom wiring? That’s the promise of MCP. MCP is an open standard that lets LLMs and third-party tools talk to each other. From hobbyists to software vendors, it makes the concept of AI agents more powerful, more useful, and way easier to extend. MCP’s co-creator, Anthropic's David Soria Parra, discusses the protocol and its promise with Yoko 👇

a16z

23,162 Aufrufe • vor 1 Jahr

Hey threat hunters and malware analysts! The web version of Matkap is now live! You can now hunt malicious Telegram bots via Happy hunting!
0:47

Hey threat hunters and malware analysts! The web version of Matkap is now live! You can now hunt malicious Telegram bots via Happy hunting!

0x6rss

62,598 Aufrufe • vor 1 Jahr