Attackers are using AI agents to automate cybercrime at 'industrial scale,' FBI and AWS warn
The FBI and AWS are warning about the increasing use of AI agents to automate cybercrime at an industrial scale.
The FBI and AWS are warning about the increasing use of AI agents to automate cybercrime at an industrial scale.
PostgreSQL has released a major update, addressing 52 vulnerabilities including a 22-year-old issue (CVE-2026-6460) that allows arbitrary code execution. The project acknowledges contributions from multiple researchers,…
PostgreSQL has released a patch to address 52 security flaws, including a 22-year-old type confusion vulnerability (CVE-2026-6460) that allows a user with object creation privileges to execute code on the server. Adminis…
As AI empowers attackers to automate attacks and enterprises integrate AI into their workflows, the ability to quickly recover from disruptions – and to restore clean data – is becoming a critical competitive advantage.…
Railway has suffered a breach resulting in $740,000 in unauthorized AI usage charges after attackers stole customer AI keys. Railway is working with AI providers to recover costs and has implemented several security meas…
The attack leveraged a third-party cross-account trust relationship, and Wiz researchers identified a campaign spanning multiple environments.
This article outlines a simple, practical threat modeling exercise for CISOs, using four key questions to identify and prioritize potential risks. Vaughn advocates for a quarterly refresh of the model and a focus on read…
Two critical vulnerabilities in PHP, CVE-2026-109618 and CVE-2026-109619, have forced 32 UK government services offline and four partially unavailable.
This article from postmortem explores how supply-chain attacks can leave a trail even after the initial damage is contained. The key takeaways are a 48-hour cooldown, a metadata check on every lockfile change, and a hunt…
AWS has since addressed the issue with updates to AgentCore, including switching to IMDSv2 and restricting agent permissions.
AWS has launched Strands Box, an open-source AI agent sandbox, to improve control over autonomous AI agents. Human oversight remains necessary, even with properly configured permissions.
A security researcher, Paulos Yibelo, discovered a ‘Full VM escape zeroday’ in Linux KVM, the hypervisor used by companies like Vercel and AWS. The Vercel Sandbox program, utilizing AWS’s Firecracker MicroVMs based on KV…
A use-after-free vulnerability in the Linux kernel's AF_UNIX socket subsystem allows an attacker to escape a container and gain root access on the host.
AWS has implemented a managed policy, AWSCompromisedKeyQuarantine, to proactively address the risk of exposed IAM access keys. This policy automatically attaches to IAM users whose credentials have been exposed, either t…
A former CrowdSec employee’s GitHub account was used to copy approximately 170 private GitHub repositories, including sensitive data like email addresses, investor details, and internal code, following a supply chain att…
Unit 42 researchers discovered a significant security vulnerability in AWS AgentCore Harness, a managed runtime for AI agents. Due to a default configuration, the built-in shell tool runs as root, allowing attackers to e…
Researchers have discovered a new hardware attack, DDRop, that breaks the memory protection offered by Intel TDX and AMD SEV-SNP, two key technologies used in cloud computing to protect customer data. The attack involves…
A significant portion of identity breaches don't start with phishing or exploits, but rather with infostealers silently collecting credentials and session data from infected machines. The Snowflake customer intrusions, w…
A recent study by Intruder found significant differences in cloud security configurations across major providers – AWS, Azure, and Google Cloud. While some issues like weak identity and access management were prevalent a…
Catch, an AI executive assistant startup, has raised $5 million to enhance its agentic admin assistant, designed to handle sensitive executive tasks like scheduling and communications while prioritizing security and user…