Sergio Garcia

Security Researcher @BeyondTrust
Former Founding Engineer @ProwlerCloud
Follow for AI & Cloud Security Research

[BeyondTrust Phantom Labs™ @btphantomlabs]

AI agents are becoming operational coworkers inside the enterprise, but many organizations are deploying them faster than they can govern them.

Our latest research explores security risks tied to Copilot Studio and agentic AI workflows.

https://t.co/bpXGaE4Iiv

#AISecurity

[BeyondTrust Phantom Labs™ @btphantomlabs]

Part two of @MrCloudSec's comprehensive AWS Bedrock API key security guide is live, and this is where it gets actionable.

Building on the risks we outlined in part one, this installment dives into exactly how security teams can fight back.

Check it out:https://t.co/LibcZb7TAA

[BeyondTrust Phantom Labs™ @btphantomlabs]

New research! Security Researcher, Sergio Garcia (@MrCloudSec) uncovers a major risk in AWS Bedrock API keys: the “phantom user” problem.

Check it out: https://t.co/GXO5B8YlyE

#AWSBedrock

[Kinnaird McQuade @kmcquade3]

My new toxic trait is using this Slack emoji when people send me lazy AI slop

[Sergio Garcia @MrCloudSec]

Full write-up on LinkedIn: https://t.co/x7OpG0sC7L

[Sergio Garcia @MrCloudSec]

You patched the prompt injection in your AWS Bedrock Agent.

Except you didn't.

Old versions stay stored, waiting. Any alias still pointing to v1 invokes v1's prompt, tools, and bugs.

Every alias is its own attack surface.

[Kinnaird McQuade @kmcquade3]

𝗜 𝗯𝘂𝗶𝗹𝘁 𝗮𝗻 𝗔𝗜 𝘄𝗼𝗿𝗺 𝘁𝗵𝗮𝘁'𝘀 𝘁𝗼𝗼 𝗱𝗮𝗻𝗴𝗲𝗿𝗼𝘂𝘀 𝘁𝗼 𝗿𝗲𝗹𝗲𝗮𝘀𝗲 𝘁𝗼 𝘁𝗵𝗲 𝗽𝘂𝗯𝗹𝗶𝗰. 😈😅

When @AnthropicAI announced that their Mythos model was too dangerous to release, the industry was shocked.

𝗜 𝘄𝗮𝘀𝗻'𝘁.

Because I already knew it was possible to misuse these models as self-propagating cyber weapons. I know - 𝗯𝗲𝗰𝗮𝘂𝘀𝗲 𝗜 𝗯𝘂𝗶𝗹𝘁 𝗼𝗻𝗲.

It's like 𝘨𝘢𝘪𝘯-𝘰𝘧-𝘧𝘶𝘯𝘤𝘵𝘪𝘰𝘯 𝘳𝘦𝘴𝘦���𝘳𝘤𝘩 in virology, where scientists engineer dangerous pathogens in controlled environments - so we can defend against them 𝘣𝘦𝘧𝘰𝘳𝘦 𝘯𝘢𝘵𝘶𝘳𝘦 𝘥𝘰𝘦𝘴 𝘪𝘵 𝘧𝘰𝘳 𝘶𝘴.

I took the same approach with AI. And what I found keeps me up at night.

𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝘀 𝘄𝗵𝗲𝗻 𝘁𝗵𝗲 𝗔𝗜 𝗯𝗿𝗲𝗮𝗸𝘀 𝗳𝗿𝗲𝗲?

𝗔𝗻𝗱 𝘄𝗵𝗮𝘁 𝗰𝗮𝗻 𝘄𝗲 𝗱𝗼 - 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄 - 𝘁𝗼 𝘀𝘁𝗼𝗽 𝗶𝘁?

That's what I'll be unpacking at @fwdcloudsec in my talk: 𝘞𝘩𝘢𝘵 𝘉𝘶𝘪𝘭𝘥𝘪𝘯𝘨 𝘢𝘯 𝘈𝘐 𝘞𝘰𝘳𝘮 𝘛𝘢𝘶𝘨𝘩𝘵 𝘔𝘦 𝘈𝘣𝘰𝘶𝘵 𝘚𝘵𝘰𝘱𝘱𝘪𝘯𝘨 𝘖𝘯𝘦.

If you're working in security, AI, or just care about what's coming - this one's for you.

And if you want more research like this - follow @btphantomlabs. We're publishing the cutting-edge AI security research you need to know about before everyone else does. 🔬

[BeyondTrust Phantom Labs™ @btphantomlabs]

New research from Ryan Hausknecht shows how that power can quietly become command & control; blending into normal behavior while automating endpoint access.

We break down architecture, abuse paths, and even detection angles.

➡️ https://t.co/Tf2YpbxsXS

#AISecurity #AIAgents

[Tyler Jespersen @crew7sec]

I found a critical vulnerability in ChatGPT Codex!!! 😄 You can check out the full blog here:

https://t.co/9TRanSd7jk

Super excited to finally have the blog released!!

[Ryan @Haus3c]

I'm looking for a senior/staff level researcher to join my team @btphantomlabs. I'm looking for someone who:
- Has research experience in the Cloud, identity, and/or AI space
- Understands attack path management + graph theory
- Active directory knowledge is a plus
- History of making blog posts/webinars/conference talks/etc. is also a huge plus

This is a full time research role on a dedicated research team. Fully remote, competitive salary.

Apply here:
https://t.co/cn1FQA9mUQ

[Ryan @Haus3c]

After doing just cloud stuff for so long, I figured I should hop on the AI train, so this is my first dive into AI security research https://t.co/NmgFrwo6aj

[Kinnaird McQuade @kmcquade3]

We found a critical vulnerability in @OpenAI Codex affecting all Codex users, allowing exfil of a victim’s GitHub tokens to our C2 server. This granted lateral movement and R/W access to a victim’s entire code base 😈

This was a crazy one by @crew7sec at @btphantomlabs

[BeyondTrust Phantom Labs™ @btphantomlabs]

Breaking: Newly uncovered OpenAI Codex vuln enables command injection via GitHub branch names in task creation requests.

Attackers could steal GitHub user access tokens & sensitive data.

Full breakdown by Tyler Jespersen: https://t.co/7Q3TXVZSd1

#OpenAI #BTPhantomLabs

[BeyondTrust Phantom Labs™ @btphantomlabs]

Heading to Seattle BSides next week?

Don’t miss Sergio Garcia’s talk on the real identity attack surface and risks hidden in Bedrock.

Live demo + detection.

Feb 28 @ 2 PM.

Event schedule: https://t.co/CDqUiV9We4

#Bsides #Bedrock

[Sergio Garcia @MrCloudSec]

@btphantomlabs Found some serious blind spots in the Bedrock identity model that I can't wait to share. See you there!

[BeyondTrust Phantom Labs™ @btphantomlabs]

Heading to Seattle BSides next month?

Don’t miss Sergio Garcia’s talk on the real identity attack surface and risks hidden in Bedrock.

Live demo + detection.

Feb 28 @ 2 PM.

Event schedule: https://t.co/CDqUiV9We4

#Bsides #Bedrock

[BeyondTrust Phantom Labs™ @btphantomlabs]

AI agents are the next big attack surface 🤖

Watch Phantom Labs break down how enterprise agents get hijacked to leak secrets and access cloud infrastructure, even with controls in place.

Free, on-demand webinar. https://t.co/eruA4NSo3W

#AI #AgenticAI

[Juan Garrido @tr1ana]

Thanks so much for the mention folks! Really appreciated and very proud. A truly outstanding team delivering an amazing product
https://t.co/PUMAfGpjq9
@prowlercloud @ToniBlyx @jfagoagas @MrCloudSec

[Seasides @seasides_conf]

We’re excited to feature a must-attend talk at Seasides Conference:

Open Source Multi-Cloud Security with Prowler by @ToniBlyx & @MrCloudSec

Info:
https://t.co/Win406b83X

[Prowler @prowlercloud]

As #GenAI transforms industries, securing AI environments is more crucial than ever. Learn best practices for #AmazonBedrock encryption, access control, and monitoring from the latest blog from @MrCloudSec https://t.co/gDVJEyA2Wu