<p><strong>TLDR:</strong> A credential-stealing worm called Miasma compromised 32 official @redhat-cloud-services npm packages on June 1, using a hijacked Red Hat employee GitHub account to push malicious GitHub Actions workflows that published backdoored versions with valid SLSA provenance. The malware sweeps for AWS, GCP, Azure, Kubernetes, Vault, SSH keys, and more, then self-propagates by republishing packages using stolen tokens. Separately, Meta’s AI support chatbot was exploited to take over Instagram accounts at scale — including the Obama White House handle — by social-engineering the bot into sending verification codes to attacker-controlled emails, bypassing 2FA entirely. Nine new <a href="http://X.Org" target="_blank" rel="noopener nofollow">X.Org</a> Server vulnerabilities were patched, eight of them found by AI-assisted research.</p>
<hr>
<h2>Miasma worm backdoors 32 Red Hat npm packages via compromised GitHub Actions pipeline</h2>
<p>On June 1, attackers compromised a Red Hat employee’s GitHub account and injected malicious orphan commits into the RedHatInsights/javascript-clients repo. The workflows abused GitHub Actions OIDC tokens to publish backdoored versions of 32 @redhat-cloud-services packages — with valid SLSA provenance attestations, so naive provenance checks won’t save you. The 4.2 MB obfuscated preinstall payload sweeps for AWS, GCP, Azure, Kubernetes, HashiCorp Vault, GitHub Actions secrets, SSH keys, Docker credentials, npm/PyPI tokens, and .env files, then self-propagates using stolen tokens to compromise additional packages. Combined weekly downloads across affected packages were roughly 80,000–117,000. Red Hat says no enterprise products were impacted due to version pinning, but any CI environment that ran npm install against affected versions should be treated as fully compromised: rotate all credentials immediately, audit GitHub for unauthorized repos and workflows, and hunt for branches named ‘chore/add-codeql-static-analysis’ as an IOC. <a href="https://app.daily.dev/feed-by-ids?id=Q3mbZ24Yd&amp;id=mpTfkMIuW&amp;id=bzd1MUxSa&amp;id=4Kacs0E4h&amp;id=jLxGALOM3&amp;id=PrnO3btx5&amp;id=IF43c4tVr&amp;id=l4Anv31dE&amp;id=Imn47QmZ8&amp;id=suwBrjHCV" target="_blank" rel="noopener nofollow">Read more</a></p>
<h2>Meta’s AI support chatbot exploited to hijack Instagram accounts, bypassing 2FA</h2>
<p>Attackers social-engineered Meta’s AI Support Assistant into adding attacker-controlled email addresses to target Instagram accounts, then used the resulting verification codes to trigger password resets — no identity verification required. A VPN was used to spoof a local IP, which appears to have been enough to make the bot treat requests as legitimate. Compromised accounts included the Obama-era White House Instagram and a US Space Force official’s account, with links to pro-Iranian groups. Meta pushed an emergency patch and confirmed the issue resolved Monday. Accounts with any MFA enabled were immune to this specific attack. The broader problem is structural: an AI chatbot with account-level write permissions operates at a scale and speed no human support team could match, and it has no intuition to flag suspicious conversation patterns. Instagram’s Trust and Safety org reportedly lost around 60% of staff through layoffs and reassignments to data labeling in the weeks prior, while aggressive AI feature pushes introduced bugs to production. <a href="https://app.daily.dev/feed-by-ids?id=ehnDbVpJ3&amp;id=dpi6fPu3x&amp;id=VtlSgFruq&amp;id=LRpaaxZaY" target="_blank" rel="noopener nofollow">Read more</a></p>
<h2>Nine <a href="http://X.Org" target="_blank" rel="noopener nofollow">X.Org</a> Server CVEs patched, eight found by AI-assisted research</h2>
<p>Patched releases xorg-server 21.1.23 and xwayland 24.1.12 fix nine vulnerabilities including stack-based buffer overflows, use-after-free bugs, and out-of-bounds read/write issues. Eight of the nine were found by Trend Micro’s TrendAI Zero Day Initiative using AI-assisted tooling — the same pattern driving the Linux AF_ALG deprecation and the PostgreSQL CVE spike from last week. Legacy codebases with decades of accumulated C are proving to be rich targets for AI-powered fuzzing and static analysis. If you’re running <a href="http://X.Org" target="_blank" rel="noopener nofollow">X.Org</a> in any production or shared environment, update now. <a href="https://app.daily.dev/posts/bvrizl9GD" target="_blank" rel="noopener nofollow">Read more</a></p>
<h2>CVE-2026-45618: CVSS 10 RCE in LiquidJS affects 7.3M monthly npm downloads</h2>
<p>A critical RCE flaw in LiquidJS, the Node.js Liquid template engine, allows attackers to reach the Function constructor via the valueOf filter and prototype manipulation, then execute arbitrary commands via child_process.execSync — no authentication required and a public PoC already exists. All versions prior to 10.26.0 are affected. With 7.3 million monthly downloads, this has a wide blast radius. Upgrade to 10.26.0 immediately; if you can’t patch, restrict template input from untrusted sources as a stopgap. <a href="https://app.daily.dev/posts/80y8SP5mi" target="_blank" rel="noopener nofollow">Read more</a></p>
<hr>
<h2>Also notable</h2>
<ul>
<li><strong>CVE-2026-41089: Windows Netlogon RCE now actively exploited in the wild:</strong> Belgium’s CCB confirmed active exploitation of this critical stack-based buffer overflow in Windows Netlogon, which allows unauthenticated RCE on domain controllers across all supported Windows Server versions including Server 2025 — Microsoft patched it in May 2026 Patch Tuesday, so if you haven’t applied it yet, that window has closed. <a href="https://app.daily.dev/posts/ZyueyB492" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palo Alto GlobalProtect CVE-2026-0257 auth bypass under active exploitation, CISA KEV listed:</strong> Rapid7 observed exploitation starting May 17 from Vultr-hosted infrastructure; the root cause is PAN-OS trusting decrypted auth override cookie contents without signature verification, and if the same certificate is shared between HTTPS services and the auth override feature — a common config — attackers can forge valid cookies and bypass VPN authentication entirely. <a href="https://app.daily.dev/posts/akuP0u3R2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CVE-2026-8732: WP Maps Pro privilege escalation actively exploited, 15,000+ sites affected:</strong> Unauthenticated attackers can create WordPress admin accounts in a single request against WP Maps Pro 6.1.0 and older via a publicly visible nonce — Defiant blocked over 3,600 exploitation attempts within 24 hours of disclosure, and because the plugin is distributed through Envato rather than the official WordPress directory, automatic updates don’t apply, so manual patching to 6.1.1 is required. <a href="https://app.daily.dev/posts/I69gHuwaM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle launches monthly patch cadence with 35 fixes, including CVSS 10 in REST Data Services:</strong> Oracle’s new ‘third Tuesday’ monthly CSPU cycle debuts with CVE-2026-46840, a CVSS 10 unauthenticated flaw in Oracle REST Data Services, plus several older vulnerabilities with public PoC exploit code — cloud customers get patches automatically, but on-premises deployments need immediate attention. <a href="https://app.daily.dev/posts/lTLKKFlSu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pwn2Own Berlin 2026: 47 zero-days, $1.3M in payouts, AI products fall on day one:</strong> OpenAI Codex, LiteLLM, LM Studio, and NVIDIA Megatron Bridge all fell to what researchers are calling the ‘trust boundary problem’ — AI tools unconditionally trusting external tools and protocols — while Microsoft Exchange yielded a SYSTEM-level RCE chain and VMware ESXi produced a cross-tenant guest-to-host escape, with AI-assisted research tools used by the majority of competing teams. <a href="https://app.daily.dev/posts/xFYxuYPgZ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>


CVE disclosures, supply-chain attacks, security incidents, and developer-facing security tooling.

Security Digest

A security-focused roundup covering several major incidents and vulnerabilities. A credential-stealing worm dubbed Miasma backdoored 32 official Red Hat npm packages via a compromised GitHub employee account, abusing GitHub Actions OIDC tokens to publish malicious versions with valid SLSA provenance; affected CI environments should rotate all credentials immediately. Meta's AI support chatbot was exploited to hijack Instagram accounts at scale — including the Obama White House handle — by social-engineering the bot into adding attacker-controlled emails and bypassing 2FA. Nine X.Org Server CVEs were patched, eight discovered via AI-assisted research. A CVSS 10 RCE in LiquidJS (7.3M monthly downloads) allows arbitrary command execution via prototype manipulation — upgrade to 10.26.0 immediately. Also covered: active exploitation of Windows Netlogon RCE, Palo Alto GlobalProtect auth bypass, WP Maps Pro privilege escalation, Oracle's new monthly patch cycle with a CVSS 10 flaw, and Pwn2Own Berlin 2026 results where AI products fell on day one.

Red Hat npm supply chain attack hits 32 packages, Instagram AI chatbot exploited for account takeover