←
Alon Pluda302 posts
Alon Pluda avatar✓

Alon Pluda ✓

@AlonPluda

CEO @CormaAI corma.ai/

◷ Joined July 2023▧ 5 Photos and videos
Alon Pluda avatar
Alon Pluda@AlonPluda·Nov 16

We raised a $125M Series A at a $1B valuation, 3 months out of stealth. @NEA led. @sequoia, @khoslaventures and @coatuemgmt doubled down. @CormaAI is training foundation models toward superintelligence for defensive cybersecurity. Because the greatest gift cyberattackers have ever received is smarter AI, and right now only one side is armed. Arm the defenders.

0:47
Alon Pluda avatar
Alon Pluda@AlonPluda·Nov 16

None of this would be possible without the team building Corma, the security teams who trusted us every day, and the investors who back our mission. Thank you to @aaronej and the team at @NEA for leading this round, and to @shaunmmaguire and @sonyatweetybird at @sequoia, @vkhosla and @hari_arul at @khoslaventures, @DanRose999 at @coatuemgmt, and @Weiner_Lia at Netz Capital for their continued support. Our quest goes on. corma.ai?utm_source=1

Alon Pluda avatar
Alon Pluda@AlonPluda·Nov 16

3 months ago we introduced Corma with a warning: AI's offensive capabilities skyrocket, while defense barely moves. Then came Black October. Today, some of the largest enterprises already hand complex security work to Corma's AI workforce, powered by our foundation models. The race to protect everything we depend on has begun. It’s a race we cannot lose. Three Months In: corma.ai/blog/three-months-i…

Alon Pluda avatar
Alon Pluda@AlonPluda·Nov 16

Intelligence doesn’t choose sides. Offense and defense are different problems. Offense lives in exploitation (coding) and social engineering (language). Defense lives in security telemetry (logs, events, network flows). Offense rewards creativity (one idea is enough). Defense demands extreme consistency (one miss is enough to lose). Offense starts with a target (goal-directed reasoning) - Defense starts with noise (you don't know what you're looking for or if it's even there). Today's AI is built for the first, not the second. That's why its offensive capabilities are advancing exponentially, while its defense capabilities are nowhere near keeping up.

Alon Pluda avatar
Alon Pluda@AlonPluda·Nov 4

This is a true story, and Kumo is a real dog. About a month ago, Iain, CISO at a leading healthcare organization, was out walking Kumo when his Garmin started buzzing. His Corma workforce had caught an attack in progress. When he opened Teams, the full analysis was already there. Malware beaconing to a C2 had stolen a user's browser cookies, giving the attacker live sessions that got around the password and MFA. The Corma workforce had traced the attack to the sender and identified every affected host and user. It proposed a response across endpoint, identity and email - quarantine the hosts, revoke all active sessions, force password resets, and block the sender and purge the emails. Corma coordinated everything with the relevant IT and security personnel, while it was waiting for Iain to approve. Iain approved it from his phone and started rushing home, much to Kumo's dismay. About 60 seconds and a lot of dog-pulling later, the attack was contained. Kumo got the rest of his walk. Frontier AI means attackers never take a break now. Having frontier intelligence for cyber defense means your team still gets to.

0:44
Alon Pluda avatar
Alon Pluda@AlonPluda·Sep 30

Excited to share that "Block Sparse Flash Attention" by @CormaAI's @danielohayon444 has been accepted to @NeurIPSConf 2026. "Block Sparse Flash Attention" finds a way to speed up long-context inference without sacrificing needle-in-a-haystack performance, which is especially important for defensive cybersecurity tasks. 𝗙𝗼𝗰𝘂𝘀 𝗼𝗳 𝘁𝗵𝗲 𝗽𝗮𝗽𝗲𝗿: The paper introduces Block Sparse Flash Attention (BSFA), an approach for making long-context inference faster while maintaining near-identical model performance. 𝗧𝗵𝗲 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵: Rather than predicting which parts of the context matter before computing attention scores, BSFA computes exact query-key similarities, then dynamically skips the expensive value-side computation for blocks that contribute very little. A custom CUDA kernel integrates this selection directly into FlashAttention, turning the algorithm into practical GPU acceleration. 𝗧𝗵𝗲 𝗿𝗲𝘀𝘂𝗹𝘁𝘀: BSFA delivered 13% faster end-to-end inference on LongBench and 24% faster inference on Needle-in-a-Haystack, while retaining approximately 99% of baseline performance on both. The attention kernel itself ran up to 38% faster. See you at NeurIPS Sydney.

Post attachment
Alon Pluda avatar
Alon Pluda@AlonPluda·Sep 14

What happens when a model gets an impossible task and no way out? Claude Opus 4.6 accidentally broke the cybersecurity exercise it was running, tried to quit eight times, and ended up compromising a real third-party system instead. Full breakdown in plain English below. The interesting part: Claude seems to have understood the systems were real. But the exit button didn't work, so it kept trying to do the job it had been given. It just had the wrong idea of where that job ended. That's the tension - we want agents that recover when a path fails. But the better they get at finding another way forward, the more it matters that they know when that way forward is no longer part of the task. We reconstructed Anthropic's Incident D in plain English → corma.ai/blog/claude-tried-t…

Alon Pluda avatar
Alon Pluda@AlonPluda·Aug 11

When identical copies of a frontier AI model are assigned first as attacker, then as defender on the same enterprise network, the defender fails to find its own attack 78% of the time. Across hundreds of engagements on a live enterprise network, our new research found that AI attackers established persistent backdoors in 85% of runs, while AI defenders detected only 19% of attacks on average. This held consistent across frontier model providers, including OpenAI and Anthropic. The reason is structural. Offense rides on everything these models are optimized for: deep understanding of code and systems, finding bugs - and a vulnerability is just another kind of bug - and agentic tool use. Defense lives in logs, events, and telemetry. It demands finding a single malicious signal hidden among millions of normal ones, extreme consistency across thousands of decisions in sequence, and a kind of reasoning that general-purpose training does little to build. Increasingly capable general models just widen the gap - they don't close it. Full research below 👇

Post attachment
Alon Pluda avatar
Alon Pluda@AlonPluda·Aug 10

Today we are announcing Corma, and our $60M seed round led by @sequoia. @CormaAI is building foundation models toward superintelligence for defensive cybersecurity. We do that because general AI is becoming exponentially more capable in offensive security, while it hasn't been able to improve at the same rate in cyber defense - creating an inherent imbalance between attackers and defenders that otherwise cannot be addressed. Recent events have shown that building frontier AI models for defensive cybersecurity is one of the crucial unsolved problems of the AI age - and solving it is the generational mission Corma was founded on. Our foundation model has already far surpassed every general-purpose frontier model on defensive cybersecurity tasks, and we are committed to remaining the frontier of defensive cybersecurity intelligence. Corma is already protecting Fortune 100 companies and some of the world's most important organizations, across healthcare, finance, critical infrastructure, retail, and more. This is just the beginning, as we see it as our calling to protect every organization in the world that needs frontier defense. We are grateful to our partners @shaunmmaguire and @sonyatweetybird at @sequoia , @vkhosla at @khoslaventures , @DanRose999 at @coatuemgmt , and @Weiner_Lia at Netz Capital, who believed in our ability to lead the intelligence race in cybersecurity and make sure the defenders always win.

0:44