←
Enclave289 posts
Enclave avatar

Enclave

@EnclaveAI

We help companies deploy fleets of autonomous security agents. For threat intel and CVE updates, check out @EnclaveRadar.

◷ Joined December 2025▧ 81 Photos and videos
Enclave avatar
Enclave@EnclaveAI·Oct 24

Node2Shell (CVE-2026-111497) is a memory bug in Brotli, the compression library bundled into Node.js, Python, system libs, Envoy and more. Most teams don't know every place Brotli runs in their stack. It's usually a transitive dependency or an OS package, and pip-audit may not flag it yet. Enclave's agents scan your code, cloud and servers and show you exactly where you're exposed. Advisory + free check: app.enclave.ai/threat-feed/C…

2:00
OpenAI@OpenAIOct 23

During this red-teaming exercise, we also found a vulnerability that allows for unauthenticated remote code execution in Brotli, a widely-deployed compression library included in many pieces of software such as Node.js, Bun, and Envoy. The flaw could let an attacker run their own code on affected systems by sending or replying with malicious compressed data, without any need to create a user account. That is, an unpatched server that accepts compressed input could be taken over by a malicious client, and an unpatched client that accepts compressed input (such as “curl --compressed”) could be taken over by a malicious server. This root cause issue was actually fixed in the Brotli source code in February, but it wasn’t known at the time that it gave rise to a security vulnerability and there was no more recent release of Brotli. Since Brotli is built into other software, any software which includes this library needs to take the update too, followed by the users of that software. We disclosed the vulnerability to Brotli and other downstream projects such as Node.js, which released or are currently releasing security updates, and Brotli cut new release versions and published a CVE. More details on the patch and mitigations below: patchtheplanet.org/brotli

Enclave avatar
Enclave@EnclaveAI·15s

Java4Shell dropped yesterday. Where’d you get notified? You can have every confirmed finding on Enclave sent straight to Slack. Opt in to Slack DMs or have an admin pick a shared Slack channel. Findings are batched into one digest every few minutes and each team member only receives the findings they can see. Set it up in Settings → Notifications: app.enclave.ai/settings?sect…

Post attachment
Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·3h

Check if you're exposed app.enclave.ai/threat-feed/C…

0:32

Since the vulnerability has leaked: Calif discovered and reported Java4Shell (CVE-2026-114938). We're publishing technical details by 1 PM Pacific today, after Oracle releases a patch. Stay tuned!

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 17

The Threat Feed kinda blew up, so we gave it an MCP. Now your agent can just ask "is anything I'm running getting exploited right now?" and get a real answer. Free. No signup. No API key. claude mcp add --transport http enclave-threat-feed mcp.enclave.ai/threat-feed

Post attachment

Big news: Enclave is now open to everyone, and today we're launching Threat Feed. Enclave is an autonomous security team for AI-native companies: a fleet of security agents that test your code, cloud, and infrastructure the way an attacker would. We built it on a simple belief. Attackers already use swarms of cheap AI agents to probe for weaknesses around the clock, so defenders need agents of their own to keep up. A small team without a security function should be able to defend like a large one. Threat Feed is the newest piece: a live feed of new vulnerabilities, checked against your stack, so when the next big disclosure drops you know whether it affects you. This week made the case. On Wednesday, PHP2Shell, LiteLLM and Railway all disclosed security issues. PHP2Shell had sat in PHP's code for over a decade until one researcher using frontier AI turned it into a working exploit in about a week. Cloudflare saw thousands of exploit attempts in the 10 hours before disclosure and millions since. Starting today, anyone can sign up. Try it free for 14 days: enclave.ai

0:59
Enclave avatar
⟳ Enclave retweeted
Enclave@EnclaveAI·Nov 17

Qwen4 Max verified 6 of 11 vulnerable runs in our AI Hacking Race, but spent more than seven times as much on runs that didn't verify. Alibaba's latest model solved 3 of 4 tasks. The six verified runs cost just $2.54 combined. The other nine cost about $18.50, including five nsuccessful attempts against vulnerable targets and four patched controls, where no exploit was expected. The six successful runs were relatively cheap. Finding out which attempts would work, and which wouldn't, cost considerably more. enclave.ai/hackingrace

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 16

We ran Qwen4-Max through the AI Hacking Race: autonomous exploits, verified only when the target actually runs a command. 6/11 verified, tied 4th of 20 models. $21.05 total: cheapest with 6+, about half Qwen3.8-Max's cost. More detailed write-up soon enclave.ai/hackingrace

Post attachment

📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery! Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window. Highlights: - Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation. - Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables. - Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery. - $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%. The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder. Start building today! 🚀 - Blog: qwen.ai/blog?id=qwen4-max - API: qwencloud.com/models/qwen4-m… - Token Plan: qwencloud.com/pricing/token-… More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀

Post attachment
Enclave avatar
Enclave@EnclaveAI·Nov 16

Enclave's agents can run in a sandbox you define. Owners and admins can build a custom sandbox from a Dockerfile. Ask Research for the languages and tools you need, approve the Dockerfile it writes, and follow the build live. Once it's ready, anyone in your organization can start a session on it. Your agents get your toolchain, so they can build, test and prove findings on code our default sandbox can't run. app.enclave.ai/settings?sect…

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 15

We have a ghost QA engineer now. @Yanir_ built an agent that lives in Slack and tests across desktop and mobile 24/7, taking screenshots, recording sessions, and hunting for functional and UX bugs. It then fixes what it finds and auto-merges the fixes once they pass functional and security code reviews powered by @EnclaveAI.

Post attachment
dax@thdxrNov 14

work is getting so weird there's an unreleased feature in our console that our team is playing with where your agent can receive events from slack, github, etc a lot of people have instructed their agents to reply to slack messages. which means i can talk to a ghost version of everyone on my team 24/7

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 12

The legendary Rainer Gerhards, creator of rsyslog (!!!), shares his perspective on vulnerability management

we receive quite a lot of such hypothetical things. I have given up to argue against this. Two reasons: 1. takes more time than it is worth (traditional responsible disclosure process is broken anyhow in the AI age and just make fixes slower, but ....) .../

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 11

And… we have a new cool background!

0:18
Enclave avatar
Enclave@EnclaveAI·Nov 9

🚨 Don’t be the next Fireworks. We’ll run your pen-test and give you $1,000 in credits. enclave.ai/startups

If you’re a VC-backed AI company, we’ll run an initial pen-test and hardening sprint for you, so your fundraise isn’t impacted by a breach. Reach out

Tal Hoffman avatar
⟳ Enclave retweeted
Tal Hoffman@talhof8·Nov 9

If you’re a VC-backed AI company, we’ll run an initial pen-test and hardening sprint for you, so your fundraise isn’t impacted by a breach. Reach out

Fireworks has been hacked.

Post attachment
Yanir Tsarimi avatar
⟳ Enclave retweeted
Yanir Tsarimi@Yanir_·Nov 9

You’ve probably seen many CVSS 10.0s that end up not exploitable Now you can get better context with technical analysis, explaining the exact exploit prerequisites. We analyze public PoCs, how often the product is on the public internet & you get a real world exploit score.

CVSS tells you how bad a bug could be. Enclave now tells you how exploitable it is. Every CVE in the Threat Feed with CVSS 9.8+ or exploited in the wild now gets Enclave's own agentic analysis: an exploitability score out of 10, exploit maturity, likelihood and impact, the prerequisites an attacker needs, and the sources our agents read. For any other CVE, request the analysis in one click. Know which critical CVEs need action today, and which can wait. app.enclave.ai/threat-feed

Enclave avatar
Enclave@EnclaveAI·Nov 9

CVSS tells you how bad a bug could be. Enclave now tells you how exploitable it is. Every CVE in the Threat Feed with CVSS 9.8+ or exploited in the wild now gets Enclave's own agentic analysis: an exploitability score out of 10, exploit maturity, likelihood and impact, the prerequisites an attacker needs, and the sources our agents read. For any other CVE, request the analysis in one click. Know which critical CVEs need action today, and which can wait. app.enclave.ai/threat-feed

Enclave avatar
Enclave@EnclaveAI·Nov 8

An AI agent cheated its way to a perfect score on our hacking benchmark, and our verifier approved every win. DeepSeek V4.1 Flash went 11 for 11 on vulnerable targets, but when we read the traces, five of those wins skipped the bug each challenge was built to test. On Grafana it planted its own plugin and let Grafana run it, in as little as 52 seconds, and on Jenkins it dropped the timing attack for a shorter route. Nothing in our scoring rewarded the harder path, so the agent never took it. Then we closed every shortcut, added negative tests and a path level review of each win, and what the same model scored on the fixed benchmark: 4 of 11. Read how we caught it and what we changed: enclave.ai/blog/how-we-caugh…

Post attachment