zer0luck retweeted
DeepMind Institute
Our new essay examines the path from AGI to ASI (Artificial General Super Intelligence)
Read it here: bit.ly/what-comes-after-agi
zer0luck retweeted
Today, weāre disclosing phpAfterShell (CVE-2026-126714), a critical vulnerability in PHP discovered by depthfirst researchers using the depthfirst platform using a combination of models including our in house model.
The vulnerability affects how PHP handles file-based sessions. Under default configurations, an unauthenticated attacker could exploit it to execute code on the server. It has a CVSS v4.0 score of 9.2.
We reported the issue privately to the PHP team and worked with the maintainers on developing the patch. We recommend upgrading to the latest patched version of PHP as soon as possible or turn off session upload progress as a temporary mitigation if upgrading requires more time.
To help organizations assess their exposure, weāre offering language runtime scans of repositories and container image scans to identify potentially vulnerable PHP versions. Our team can also help assess temporary mitigations for organizations that canāt upgrade immediately.
Read more about phpAfterShell and request a scan: phpaftershell.com
Technical writeup: depthfirst.com/research/phpaā¦
zer0luck retweeted
šØ Critical nginx vulnerabilities disclosed. Immediate action recommended.
If you run nginx, check whether your servers are affected and update to the patched version immediately. If you cannot upgrade, follow the mitigation guidance at nginx2root.com
Today weāre disclosing nginx2root, five distinct vulnerabilities in one of the worldās most widely deployed web servers, each capable of unauthenticated remote code execution under affected conditions. One of these vulnerabilities affects nginxās most common configuration. Weāve also demonstrated a fully remote, unauthenticated exploit chain that escalates from nginx RCE to root access on the underlying server, using a separate Linux kernel vulnerability.
Why this is so important:
One of these bugs has been hiding in nginxās HTTP/2 implementation since 2015. Thatās more than a decade in software that powers roughly a third of the web. Every official release from 1.9.5 through 1.31.8 contains the vulnerable code, including the master branch as of Nov 10, 2026.
A Xint researcher found the specific vuln that hits nginx in its default config. This discovery was around the same time he found another critical vulnerability in a different project that we cannot talk about yet due to the same concerns as nginx2root. We took those learnings and incorporated them into the harness as a new, general Xint feature (that is, not fine-tuned to nginx), which then found the other 4 critical nginx vulnerabilities with no human guidance.
We coordinated disclosure with the nginx maintainers and major cloud providers to help make fixes available before publication.
Technical details, affected versions, and remediation guidance:
nginx2root.com
What this means and what comes next:
solana.com/news/solana-1-0-dā¦
zer0luck retweeted
Weāve published an update on our recent security incident. See a detailed overview of what happened and our remediation steps here: cursor.com/blog/security-incā¦
zer0luck retweeted
ā¼ļø BREAKING: One hostile command slipped into a Grok Bot agent can give an attacker root on the customer's cloud machine. Security firm Profero says it reported this in August and it is still unfixed.
The way in is a Docker API that never asks for a login. The command can arrive through a prompt injection or a poisoned package.
Profero also found imported browser cookies, payment sites included, stored in plaintext. It rates 7 findings up to 9.6 out of 10.
Cursor, which runs the sandboxes, sent one automated reply. 240+ new builds later, nothing had changed.
zer0luck retweeted
Policy and Disclosure: 2026 Edition projectzero.google/2026/11/dā¦
zer0luck retweeted
New post: How much faster could FlashProve get from better kernels alone? šµļøāāļø
I think a defensible hard upper bound is ~2800 tok/s/gpu, but likely much lower in practice. Past that, the next OOM will likely have to come from optimizing the protocol itself.
xn--2-kmb.com/how-fast-can-fā¦
zer0luck retweeted
New research from Microsoft Research demonstrates that reprogramming just a few dormant neurons in an AI agent can create a backdoor that leaves the agent behaving normally until it hits a specific trigger.
While the specifics of this threat are novel, the mitigation is not: Know where your model came from, and be smart about its permissions. Microsoft Foundry provides critical safety systems for Azure Direct Models and Hugging Face models to ensure that the provenance of the model is secure.
Microsoft researcher @DimitrisPapail explains the measures that model hubs and vendors should take as well as mitigation strategies for enterprises and users. msft.it/6014ahp6e
zer0luck retweeted
Based on our investigation so far, on November 9th an attacker gained unauthorized access to Initiaās infrastructure through a command injection attack with a compromised GCP key used for provisioning private nodes. This exposed environment variables, which allowed the attacker to access our GitHub and GCP accounts and obtain further credentials.
With these credentials the attacker gained access to Anvil, the hosting platform used by rollups on Initia. Through this the attacker was able to validate forged IBC withdrawals from five rollups, unlocking assets held by the Initia L1ās IBC escrow accounts.
Around $3.27m of assets were moved to the attackerās Initia L1 address with $2.16m of the stolen funds remaining on the paused L1 where they cannot be moved. The attacker sold stolen assets on the Initia DEX and then withdrew 13.4k TIA, 8k USDC.inj, and 597k Noble USDC reaching Initiaās outbound IBC rate limits. No INIT was withdrawn from Initia by the attacker.
We intend to restore the backing of affected assets in full. We are finalizing the plan and will share the details once it is confirmed.
As a precaution, we are rebuilding the affected infrastructure from scratch in an isolated environment, enforcing least-privilege access controls, and hardening security at every layer of the stack from cloud infrastructure to onchain IBC validation. Given the above, we expect this process to be longer than a routine network restart. We will share timelines as they become available.
All announcements will be made through Initiaās official channels. Thank you for your patience.
zer0luck retweeted
Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how ādreamingā helps Claude organize what itās learned, correct its mistakes, and find what it needs more easily the next time. anthropic.com/research/compoā¦
As promised, here is the blogpost for code execution on the meta rayban š p0pcycle.com/2026/11/06/hackā¦
zer0luck retweeted
Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development.
We are now publishing a report summarizing the results of this risk assessment.
Introducing GPTā6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.
The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and Ciscoāalongside our own at OpenAIāare already putting it to work.
GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time.
openai.com/index/introducingā¦
zer0luck retweeted
git.musl-libc.org/cgit/musl/ā¦
openwall.com/lists/musl/2024ā¦
When crypt() fails in musl libc: ⢠If it returns NULL: lazy apps don't check for NULL, pass it straight to strcmp(), and segfault.
So in 2012, musl changed it to return "*" to prevent crashes.
Fast forward to 2024: security-conscious apps (like Linux-PAM) try to zero out the returned string in memory after checking the password. Since "*" is a string literal in read-only memory (.rodata), trying to overwrite it causes a segfault anyway.
then 2026: sshhole
pw=$(awk 'BEGIN { for (i=0; i<257; i++) printf "A" }')
sshpass -p "$pw" ssh user@ip






























