Justin Elze
RT @orange_8361: Bonus: I've also flipped the chain and popped MS Edge from calc.exe 😆
082096
Justin Elze✓
@HackingLZ
CTO @TrustedSec | Former Optiv/SecureWorks/Accuvant Labs/Redspin | Race cars
5K Following75.5K Followers67.9K Posts
- Justin ElzeRT @thdxr: ok to put what i'm saying about demand in context openrouter is the most successful model router, you can almost serve 100% of their deepseek traffic yesterday on a single GB300 rack this is worth 6-7M...even if there's 10x this demand secretly that's still < 100M of compute01600104
- Justin ElzeRT @MRatable: High priority: AI Firewall AI Endpoint Protection AI SIEM AI Identity Low priority: Firewall Endpoint Protection SIEM Identitya16z @a16zIn a single year, AI has rewritten the security priority list for CISOs. Attackers are moving faster. Bain estimates AI compresses a typical cyberattack from about four weeks to about 18 hours, and models keep getting better at finding and exploiting weaknesses. There's also more to protect, with employees shipping apps they couldn't have built a year ago and agents getting access to source code, documents, and production systems. In a survey of 100 CISOs at companies with 1,000+ employees, all seven of the most-selected 2027 priorities involve AI. The biggest jumps since 2026: - Internal AI agent security (+28 pts) - AI SOC (+26 pts) - AI offensive security (+24 pts) - AI code security (+23 pts) - AI vulnerability remediation (+22 pts) Last year's staples: - Exposure & vulnerability management (-21 pts) - SIEM / security data (-18 pts) - Privileged access management (-18 pts) From a16z's @OwenColby and @steph_zhang on The New Security Agenda:0150146
a16z@a16zIn a single year, AI has rewritten the security priority list for CISOs. Attackers are moving faster. Bain estimates AI compresses a typical cyberattack from about four weeks to about 18 hours, and models keep getting better at finding and exploiting weaknesses. There's also more to protect, with employees shipping apps they couldn't have built a year ago and agents getting access to source code, documents, and production systems. In a survey of 100 CISOs at companies with 1,000+ employees, all seven of the most-selected 2027 priorities involve AI. The biggest jumps since 2026: - Internal AI agent security (+28 pts) - AI SOC (+26 pts) - AI offensive security (+24 pts) - AI code security (+23 pts) - AI vulnerability remediation (+22 pts) Last year's staples: - Exposure & vulnerability management (-21 pts) - SIEM / security data (-18 pts) - Privileged access management (-18 pts) From a16z's @OwenColby and @steph_zhang on The New Security Agenda:
64106696105K- Ha "solved for hardened targets, and the remaining gap is the un-hardened default"0091.7K
- Justin ElzeRT @gregce10: You would not even believe the amount of credentials / keys (aws, gcp, azure, saas vendors, oai, ant, etc) exposed to models via agents on a daily basis in agent session logs. I don't even think you can even fathom the volume of exposure.020134
- This part is fun because while enterprises trust models for coding they don’t in a lot of other areas.
3381.7K - Posts @HackingLZI spend a crazy amount of time each week talking about frontier vs local and various levels of human oversight/gating. Each random sandbox escape news story sets this back as well.100533
- Justin ElzeRT @_rybaz: Favorite part of talking with @lcfr_hax at @unpromptedconf is learning that he didn't do anything crazy with harnessing. He's just really fucking good at hacking PHP and also happens to have Codex. Such a friendly and humble dude. Don't just "use AI". Keep learning.070135
- Everything shouldn’t be a tweetTony Chang ☁️ @tonychang430An engineer on our team used roughly $10,000 in @claudeai API credits in a MONTH. So we moved to using Claude Team subscriptions instead at roughly $100–$200 per seat, with each of us using 3 - 7 subscriptions. This doesn't make any sense 💀5123587K
An engineer on our team used roughly $10,000 in @claudeai API credits in a MONTH. So we moved to using Claude Team subscriptions instead at roughly $100–$200 per seat, with each of us using 3 - 7 subscriptions. This doesn't make any sense 💀713623214K- Justin ElzeRT @PyroTek3: TIL: Microsoft Copilot keeps conversation history and personalization memory, stored in the user’s Exchange mailbox, so retention and eDiscovery apply. Prompt history, data artifacts, and memory data are stored in hidden folders in the user's Exchange online mailbox, similar to Teams messages. This means that an attacker with full access to mailboxes can pull this information.0480115
- Justin ElzeRT @Guluisacat: Planned to drop a macOS 0day at OffensiveCon 2026 in Berlin 5 months ago. Submitted it to Apple 3 years ago, and Apple confirmed the vulnerability but closed the report. it's a universal method to bypass many security protections on macOS and you can fully bypass TCC and gain other privileged access as you want, easilly weaponized. Although I mentioned the impact multiple times, there were no updates for the next 2 years until my presentation was accepted by OffensiveCon26 Berlin. Apple asked me to hold off on disclosure and not disclose the vulnerability at OffensiveCon26 Berlin, saying a patch would be released in macOS 27 in Fall 2026. They also mentioned "bounty" multiple times. Honestly, I did have some complaints because their workflow was really bad. No updates at all until disclosure was about to happen. But as long as they decided to patch it, I was willing to wait, even without a bug bounty, for user safety. If Apple wasn't going to release a patch for 3 years, I'd disclose the vulnerability so EDR vendors could protect users too. But in September, when the vulnerability was patched, there was no cve, no credit, and no bounty. I really don't get why apple chose to handle it in such a dishonest and underhanded way, a cheat. Even if they had never mentioned a bounty, I still would have held off on disclosure because they were going to publish a security bulletin. this is a huge disappointment in my experience with the Apple Security Bounty program. never thought I'd encounter something like this. As i said on my blog a few years ago, my experience with ASB had been positive. that's why I was willing to wait for the patch even without a bounty, but not to be cheated like this01280132
- Justin ElzeRT @v12sec: We found bugs in the AES-256 encryption of multiple PDF editors. Some so weak you could break them by hand 🫠0210146
- Justin ElzeRT @KimZetter: For years government officials have been telling us how sophisticated and stealth China's Volt Typhoon hackers are, without providing details that support this assertion. Today researchers at @BlackLotusLabs finally reveal details that back the claim0920151
- Vulnerability management eh?a16z @a16zIn a single year, AI has rewritten the security priority list for CISOs. Attackers are moving faster. Bain estimates AI compresses a typical cyberattack from about four weeks to about 18 hours, and models keep getting better at finding and exploiting weaknesses. There's also more to protect, with employees shipping apps they couldn't have built a year ago and agents getting access to source code, documents, and production systems. In a survey of 100 CISOs at companies with 1,000+ employees, all seven of the most-selected 2027 priorities involve AI. The biggest jumps since 2026: - Internal AI agent security (+28 pts) - AI SOC (+26 pts) - AI offensive security (+24 pts) - AI code security (+23 pts) - AI vulnerability remediation (+22 pts) Last year's staples: - Exposure & vulnerability management (-21 pts) - SIEM / security data (-18 pts) - Privileged access management (-18 pts) From a16z's @OwenColby and @steph_zhang on The New Security Agenda:20715716.9K
- Show more posts
