Nikoloz K.
@nikolozk.bsky.social
Founder of CybersecTools.com (3,000+ security tools) | Fractional CISO for B2B companies | ex-Mambu, ex-EclecticIQ, ex-JDE
AI Companies are allowing everyone to install unverified code, and no one is stopping them.
Figma's MCP tool has just had a serious security issue that allowed hackers to execute code remotely.
New MCPs are released daily, but AI companies fail to verify their safety before they are used widely.
Figma's MCP tool has just had a serious security issue that allowed hackers to execute code remotely.
New MCPs are released daily, but AI companies fail to verify their safety before they are used widely.
October 15, 2025 at 9:58 AM
AI Companies are allowing everyone to install unverified code, and no one is stopping them.
Figma's MCP tool has just had a serious security issue that allowed hackers to execute code remotely.
New MCPs are released daily, but AI companies fail to verify their safety before they are used widely.
Figma's MCP tool has just had a serious security issue that allowed hackers to execute code remotely.
New MCPs are released daily, but AI companies fail to verify their safety before they are used widely.
I just shipped 240+ hours of work into the biggest CybersecTools update yet. Here's what actually changed:
🎯 FOR SECURITY TEAMS
→ Find tools in seconds, not hours
18 categories → 106 specializations → 944 specific tasks
Need "API security testing for cloud"? Go straight there.
🎯 FOR SECURITY TEAMS
→ Find tools in seconds, not hours
18 categories → 106 specializations → 944 specific tasks
Need "API security testing for cloud"? Go straight there.
October 13, 2025 at 12:54 PM
I just shipped 240+ hours of work into the biggest CybersecTools update yet. Here's what actually changed:
🎯 FOR SECURITY TEAMS
→ Find tools in seconds, not hours
18 categories → 106 specializations → 944 specific tasks
Need "API security testing for cloud"? Go straight there.
🎯 FOR SECURITY TEAMS
→ Find tools in seconds, not hours
18 categories → 106 specializations → 944 specific tasks
Need "API security testing for cloud"? Go straight there.
Another day, another breach: this time, it's Red Hat.
And yes, their own on-premises GitLab instance.
Shocking? Not really.
Today, it doesn't really matter if you host your systems on-premises or in the cloud.
And yes, their own on-premises GitLab instance.
Shocking? Not really.
Today, it doesn't really matter if you host your systems on-premises or in the cloud.
October 9, 2025 at 1:04 PM
Another day, another breach: this time, it's Red Hat.
And yes, their own on-premises GitLab instance.
Shocking? Not really.
Today, it doesn't really matter if you host your systems on-premises or in the cloud.
And yes, their own on-premises GitLab instance.
Shocking? Not really.
Today, it doesn't really matter if you host your systems on-premises or in the cloud.
I use MCPs all the time, especially with Claude Code.
They’re game-changers for building and extending LLMs.
But let’s be honest: from a cybersecurity perspective, MCPs are a dangerous wild card for enterprises.
Right now, it’s almost impossible to verify which MCP servers are legit.
They’re game-changers for building and extending LLMs.
But let’s be honest: from a cybersecurity perspective, MCPs are a dangerous wild card for enterprises.
Right now, it’s almost impossible to verify which MCP servers are legit.
October 8, 2025 at 12:58 PM
I use MCPs all the time, especially with Claude Code.
They’re game-changers for building and extending LLMs.
But let’s be honest: from a cybersecurity perspective, MCPs are a dangerous wild card for enterprises.
Right now, it’s almost impossible to verify which MCP servers are legit.
They’re game-changers for building and extending LLMs.
But let’s be honest: from a cybersecurity perspective, MCPs are a dangerous wild card for enterprises.
Right now, it’s almost impossible to verify which MCP servers are legit.
Qantas cuts CEO pay by $800,000 after a data breach.
Many are calling this a “win” for CISOs.
Regulators are preparing to fine executives directly when security breaches occur.
People are calling this an 'accountability revolution' in cybersecurity, but I think that's premature.
Many are calling this a “win” for CISOs.
Regulators are preparing to fine executives directly when security breaches occur.
People are calling this an 'accountability revolution' in cybersecurity, but I think that's premature.
September 30, 2025 at 12:53 PM
Qantas cuts CEO pay by $800,000 after a data breach.
Many are calling this a “win” for CISOs.
Regulators are preparing to fine executives directly when security breaches occur.
People are calling this an 'accountability revolution' in cybersecurity, but I think that's premature.
Many are calling this a “win” for CISOs.
Regulators are preparing to fine executives directly when security breaches occur.
People are calling this an 'accountability revolution' in cybersecurity, but I think that's premature.
CyberSecTools has just reached 885 registered users and catalogued 3,005 security tools.
In just the past 28 days:
→ 13,500 unique visitors
→ 8,500 from organic search engines
→ 100+ visitors from ChatGPT
ChatGPT is now sending us traffic as people ask it for cybersecurity tool recommendations.
In just the past 28 days:
→ 13,500 unique visitors
→ 8,500 from organic search engines
→ 100+ visitors from ChatGPT
ChatGPT is now sending us traffic as people ask it for cybersecurity tool recommendations.
September 12, 2025 at 6:00 AM
CyberSecTools has just reached 885 registered users and catalogued 3,005 security tools.
In just the past 28 days:
→ 13,500 unique visitors
→ 8,500 from organic search engines
→ 100+ visitors from ChatGPT
ChatGPT is now sending us traffic as people ask it for cybersecurity tool recommendations.
In just the past 28 days:
→ 13,500 unique visitors
→ 8,500 from organic search engines
→ 100+ visitors from ChatGPT
ChatGPT is now sending us traffic as people ask it for cybersecurity tool recommendations.
AWS just released Strands Agents – an open source SDK that revolutionizes how we build AI agents with minimal code.
The traditional approach to AI agent development has been painfully complex – requiring intricate orchestration, custom parsers, and months of tuning.
aws.amazon.com/blogs/openso...
The traditional approach to AI agent development has been painfully complex – requiring intricate orchestration, custom parsers, and months of tuning.
aws.amazon.com/blogs/openso...
May 21, 2025 at 12:53 PM
AWS just released Strands Agents – an open source SDK that revolutionizes how we build AI agents with minimal code.
The traditional approach to AI agent development has been painfully complex – requiring intricate orchestration, custom parsers, and months of tuning.
aws.amazon.com/blogs/openso...
The traditional approach to AI agent development has been painfully complex – requiring intricate orchestration, custom parsers, and months of tuning.
aws.amazon.com/blogs/openso...
While 89% of organizations discuss security at board level, exposure management ranks LAST in investment priorities.
The data tells a sobering story:
‣ 55% of organizations still operate with crippling IT/security data silos
‣ 62% say these silos slow security response times
The data tells a sobering story:
‣ 55% of organizations still operate with crippling IT/security data silos
‣ 62% say these silos slow security response times
May 2, 2025 at 1:05 PM
While 89% of organizations discuss security at board level, exposure management ranks LAST in investment priorities.
The data tells a sobering story:
‣ 55% of organizations still operate with crippling IT/security data silos
‣ 62% say these silos slow security response times
The data tells a sobering story:
‣ 55% of organizations still operate with crippling IT/security data silos
‣ 62% say these silos slow security response times
10 books that transformed my cybersecurity leadership journey and life in ways no certification ever could.
Here's my decade-tested list:
Want to cultivate a strategic security mindset? → Join 1000+ cybersecurity leaders developing executive-level thinking: mandos.io/newsletter
Here's my decade-tested list:
Want to cultivate a strategic security mindset? → Join 1000+ cybersecurity leaders developing executive-level thinking: mandos.io/newsletter
May 1, 2025 at 1:01 PM
10 books that transformed my cybersecurity leadership journey and life in ways no certification ever could.
Here's my decade-tested list:
Want to cultivate a strategic security mindset? → Join 1000+ cybersecurity leaders developing executive-level thinking: mandos.io/newsletter
Here's my decade-tested list:
Want to cultivate a strategic security mindset? → Join 1000+ cybersecurity leaders developing executive-level thinking: mandos.io/newsletter
Are LLMs Actually Reliable for Cyber Threat Intelligence?
Research just published shows LLMs failing dramatically on actual threat reports. Previous papers claimed 80-90% accuracy, but they tested on tiny text samples (fewer than 200 words). Real CTI reports average 3,000+ words.
Research just published shows LLMs failing dramatically on actual threat reports. Previous papers claimed 80-90% accuracy, but they tested on tiny text samples (fewer than 200 words). Real CTI reports average 3,000+ words.
April 27, 2025 at 2:00 PM
Are LLMs Actually Reliable for Cyber Threat Intelligence?
Research just published shows LLMs failing dramatically on actual threat reports. Previous papers claimed 80-90% accuracy, but they tested on tiny text samples (fewer than 200 words). Real CTI reports average 3,000+ words.
Research just published shows LLMs failing dramatically on actual threat reports. Previous papers claimed 80-90% accuracy, but they tested on tiny text samples (fewer than 200 words). Real CTI reports average 3,000+ words.
I analyzed Stanford's 2025 AI Index Report.
These 10 findings go beyond the headlines:
AI is advancing at unprecedented speed, transforming business, science, and daily life.
After diving deep into Stanford's comprehensive report, these key trends emerge with clarity.
These 10 findings go beyond the headlines:
AI is advancing at unprecedented speed, transforming business, science, and daily life.
After diving deep into Stanford's comprehensive report, these key trends emerge with clarity.
April 25, 2025 at 1:05 PM
I analyzed Stanford's 2025 AI Index Report.
These 10 findings go beyond the headlines:
AI is advancing at unprecedented speed, transforming business, science, and daily life.
After diving deep into Stanford's comprehensive report, these key trends emerge with clarity.
These 10 findings go beyond the headlines:
AI is advancing at unprecedented speed, transforming business, science, and daily life.
After diving deep into Stanford's comprehensive report, these key trends emerge with clarity.
Perfect security is the easy part.
Want maximum security? Disconnect everything and lock it in a vault.
Congratulations, you've achieved 99% security and 0% business value.
The real cybersecurity challenge isn't building Fort Knox. It's about:
Want maximum security? Disconnect everything and lock it in a vault.
Congratulations, you've achieved 99% security and 0% business value.
The real cybersecurity challenge isn't building Fort Knox. It's about:
April 24, 2025 at 1:01 PM
Perfect security is the easy part.
Want maximum security? Disconnect everything and lock it in a vault.
Congratulations, you've achieved 99% security and 0% business value.
The real cybersecurity challenge isn't building Fort Knox. It's about:
Want maximum security? Disconnect everything and lock it in a vault.
Congratulations, you've achieved 99% security and 0% business value.
The real cybersecurity challenge isn't building Fort Knox. It's about:
𝗪𝗶𝘇 𝗠𝗖𝗣 𝘃𝘀. 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻𝘀
Yesterday Wiz announced their new MCP Server - an implementation of the Model Context Protocol that's reshaping how AI and security tools communicate.
The security community is buzzing about it, but what's the real impact?
Let's break it down.
Yesterday Wiz announced their new MCP Server - an implementation of the Model Context Protocol that's reshaping how AI and security tools communicate.
The security community is buzzing about it, but what's the real impact?
Let's break it down.
April 22, 2025 at 12:44 PM
𝗪𝗶𝘇 𝗠𝗖𝗣 𝘃𝘀. 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻𝘀
Yesterday Wiz announced their new MCP Server - an implementation of the Model Context Protocol that's reshaping how AI and security tools communicate.
The security community is buzzing about it, but what's the real impact?
Let's break it down.
Yesterday Wiz announced their new MCP Server - an implementation of the Model Context Protocol that's reshaping how AI and security tools communicate.
The security community is buzzing about it, but what's the real impact?
Let's break it down.
𝗦𝗲𝗰𝘂𝗿𝗶𝗻𝗴 𝗔𝗜-𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱 𝗖𝗼𝗱𝗲 𝘄𝗶𝘁𝗵 𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗦𝗲𝗹𝗳-𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀
CrowdStrike just revealed a fascinating proof of concept at NVIDIA GTC 2025 - using AI to secure AI-generated code.
With "vibe coding" becoming mainstream, we're facing a potential flood of vulnerable code.
CrowdStrike just revealed a fascinating proof of concept at NVIDIA GTC 2025 - using AI to secure AI-generated code.
With "vibe coding" becoming mainstream, we're facing a potential flood of vulnerable code.
April 20, 2025 at 3:53 PM
𝗦𝗲𝗰𝘂𝗿𝗶𝗻𝗴 𝗔𝗜-𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗲𝗱 𝗖𝗼𝗱𝗲 𝘄𝗶𝘁𝗵 𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝗦𝗲𝗹𝗳-𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀
CrowdStrike just revealed a fascinating proof of concept at NVIDIA GTC 2025 - using AI to secure AI-generated code.
With "vibe coding" becoming mainstream, we're facing a potential flood of vulnerable code.
CrowdStrike just revealed a fascinating proof of concept at NVIDIA GTC 2025 - using AI to secure AI-generated code.
With "vibe coding" becoming mainstream, we're facing a potential flood of vulnerable code.
I recently reviewed Microsoft's AI Security Guide on securing AI applications.
After digging through the data, here are 5 findings worth sharing:
After digging through the data, here are 5 findings worth sharing:
April 18, 2025 at 6:58 AM
I recently reviewed Microsoft's AI Security Guide on securing AI applications.
After digging through the data, here are 5 findings worth sharing:
After digging through the data, here are 5 findings worth sharing:
Most organizations exist in a dangerous gap: Level 4-5 traditional security but Level 1 AI security maturity. This creates a false sense of security that leaves AI systems completely exposed.
Your robust SOC won't detect:
Your robust SOC won't detect:
April 16, 2025 at 9:12 AM
Most organizations exist in a dangerous gap: Level 4-5 traditional security but Level 1 AI security maturity. This creates a false sense of security that leaves AI systems completely exposed.
Your robust SOC won't detect:
Your robust SOC won't detect:
𝗢𝗽𝗲𝗻𝘀𝗼𝘂𝗿𝗰𝗲 𝗠𝗖𝗣 𝗦𝗲𝗿𝘃𝗲𝗿𝘀 𝗳𝗼𝗿 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗧𝗲𝗮𝗺𝘀
A GitHub repository of 300 open-source MCP servers just dropped, including 16 security-focused implementations that could transform your SecOps workflow.
But what does this mean for security professionals?
(continue reading for a link)
A GitHub repository of 300 open-source MCP servers just dropped, including 16 security-focused implementations that could transform your SecOps workflow.
But what does this mean for security professionals?
(continue reading for a link)
April 13, 2025 at 12:34 PM
𝗢𝗽𝗲𝗻𝘀𝗼𝘂𝗿𝗰𝗲 𝗠𝗖𝗣 𝗦𝗲𝗿𝘃𝗲𝗿𝘀 𝗳𝗼𝗿 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗧𝗲𝗮𝗺𝘀
A GitHub repository of 300 open-source MCP servers just dropped, including 16 security-focused implementations that could transform your SecOps workflow.
But what does this mean for security professionals?
(continue reading for a link)
A GitHub repository of 300 open-source MCP servers just dropped, including 16 security-focused implementations that could transform your SecOps workflow.
But what does this mean for security professionals?
(continue reading for a link)
7 Realizations about Trump's Executive Orders on AI
1. Shared knowledge repositories are game-changing. M-25-21 and M-25-22 mandate government-wide sharing of AI best practices, creating the first collaborative AI knowledge ecosystem across federal agencies.
1. Shared knowledge repositories are game-changing. M-25-21 and M-25-22 mandate government-wide sharing of AI best practices, creating the first collaborative AI knowledge ecosystem across federal agencies.
April 12, 2025 at 3:57 PM
7 Realizations about Trump's Executive Orders on AI
1. Shared knowledge repositories are game-changing. M-25-21 and M-25-22 mandate government-wide sharing of AI best practices, creating the first collaborative AI knowledge ecosystem across federal agencies.
1. Shared knowledge repositories are game-changing. M-25-21 and M-25-22 mandate government-wide sharing of AI best practices, creating the first collaborative AI knowledge ecosystem across federal agencies.
Google's new Agent-to-Agent (A2A) protocol just launched, and it's quietly transforming how AI agents work together. While everyone debates hallucinations, the real revolution is happening in agent interoperability.
April 10, 2025 at 2:45 PM
Google's new Agent-to-Agent (A2A) protocol just launched, and it's quietly transforming how AI agents work together. While everyone debates hallucinations, the real revolution is happening in agent interoperability.
Model Context Protocol (MCP) is the new Wild West of AI security risks. Are you prepared?
The Model Context Protocol (MCP) is revolutionizing how AI applications access data and tools - essentially becoming the "USB-C port for AI." But this powerful capability creates serious security blind spots.
The Model Context Protocol (MCP) is revolutionizing how AI applications access data and tools - essentially becoming the "USB-C port for AI." But this powerful capability creates serious security blind spots.
April 9, 2025 at 9:11 AM
Model Context Protocol (MCP) is the new Wild West of AI security risks. Are you prepared?
The Model Context Protocol (MCP) is revolutionizing how AI applications access data and tools - essentially becoming the "USB-C port for AI." But this powerful capability creates serious security blind spots.
The Model Context Protocol (MCP) is revolutionizing how AI applications access data and tools - essentially becoming the "USB-C port for AI." But this powerful capability creates serious security blind spots.
Your cybersecurity resume is probably sabotaging your career.
Just came across an interesting case of how Ankit Masrani landed a cybersecurity role at Microsoft without a traditional security background. See his resume:
Just came across an interesting case of how Ankit Masrani landed a cybersecurity role at Microsoft without a traditional security background. See his resume:
March 31, 2025 at 6:54 AM
Your cybersecurity resume is probably sabotaging your career.
Just came across an interesting case of how Ankit Masrani landed a cybersecurity role at Microsoft without a traditional security background. See his resume:
Just came across an interesting case of how Ankit Masrani landed a cybersecurity role at Microsoft without a traditional security background. See his resume:
AI in DLP is no longer a buzzword. It's finally solving our biggest pain point: false positives.
Cloudflare just announced an AI-powered context analysis feature for their DLP solution that actually learns from your feedback to reduce false positives over time.
As someone who's battled DLP alert f
Cloudflare just announced an AI-powered context analysis feature for their DLP solution that actually learns from your feedback to reduce false positives over time.
As someone who's battled DLP alert f
March 28, 2025 at 7:00 AM
AI in DLP is no longer a buzzword. It's finally solving our biggest pain point: false positives.
Cloudflare just announced an AI-powered context analysis feature for their DLP solution that actually learns from your feedback to reduce false positives over time.
As someone who's battled DLP alert f
Cloudflare just announced an AI-powered context analysis feature for their DLP solution that actually learns from your feedback to reduce false positives over time.
As someone who's battled DLP alert f
Anthropic's breakthrough paper reveals Claude's inner workings:
‣ Universal "language of thought" across languages
‣ Plans ahead in writing
‣ Can fabricate reasoning when misled
‣ Declines answering when unsure
Useful? Join my newsletter for more: mandos.io/newsletter
‣ Universal "language of thought" across languages
‣ Plans ahead in writing
‣ Can fabricate reasoning when misled
‣ Declines answering when unsure
Useful? Join my newsletter for more: mandos.io/newsletter
March 27, 2025 at 5:33 PM
Anthropic's breakthrough paper reveals Claude's inner workings:
‣ Universal "language of thought" across languages
‣ Plans ahead in writing
‣ Can fabricate reasoning when misled
‣ Declines answering when unsure
Useful? Join my newsletter for more: mandos.io/newsletter
‣ Universal "language of thought" across languages
‣ Plans ahead in writing
‣ Can fabricate reasoning when misled
‣ Declines answering when unsure
Useful? Join my newsletter for more: mandos.io/newsletter
Utility-based AI agents: The risk managers your security team desperately needs.
Let's talk about the cybersecurity equivalent of emergency room triage.
Unlike goal-based agents that see the world in black and white, utility-based agents operate on a spectrum of preferences.
Let's talk about the cybersecurity equivalent of emergency room triage.
Unlike goal-based agents that see the world in black and white, utility-based agents operate on a spectrum of preferences.
March 26, 2025 at 8:00 AM
Utility-based AI agents: The risk managers your security team desperately needs.
Let's talk about the cybersecurity equivalent of emergency room triage.
Unlike goal-based agents that see the world in black and white, utility-based agents operate on a spectrum of preferences.
Let's talk about the cybersecurity equivalent of emergency room triage.
Unlike goal-based agents that see the world in black and white, utility-based agents operate on a spectrum of preferences.
Your VPC Endpoints are a security blind spot. AWS just fixed that.
AWS released CloudTrail network activity events for VPC Endpoints, and it's a game-changer for those of us battling the data exfiltration problem.
AWS released CloudTrail network activity events for VPC Endpoints, and it's a game-changer for those of us battling the data exfiltration problem.
March 25, 2025 at 8:00 AM
Your VPC Endpoints are a security blind spot. AWS just fixed that.
AWS released CloudTrail network activity events for VPC Endpoints, and it's a game-changer for those of us battling the data exfiltration problem.
AWS released CloudTrail network activity events for VPC Endpoints, and it's a game-changer for those of us battling the data exfiltration problem.