Post-Quantum Rollouts, Nanosecond Sync, And Active Exploitation

Post-quantum protections expanded across a SASE stack, as outlined in a Cloudflare blog, while a software-first clock sync system, Firefly, aims for nanosecond alignment on commodity hardware. In the console, generative AI now renders linked, query-driven resource and cost views via Amazon Q. Active threats continued as developer supply chains, email platforms, and mobile device managers faced targeted exploitation.

Post‑Quantum And Precision Timing Land In Production

Cloudflare detailed standards‑compliant post‑quantum cryptography across its Secure Web Gateway, Zero Trust, and WAN services, mixing ML‑KEM with classical Diffie‑Hellman over TLS, MASQUE, and IPsec. The Cloudflare One Appliance added PQ support in version 2026.2.0 (GA), and IPsec’s IKEv2 Responder implements draft‑ietf‑ipsecme‑ikev2‑mlkem in a closed beta. The company frames key‑establishment migration as the pragmatic first step while PQ signatures remain a later move, citing NIST’s 2030 timeline and CISA’s January 2026 guidance. Why it matters: hybrid ML‑KEM helps counter “harvest‑now, decrypt‑later” risks without specialized hardware or ciphersuite sprawl.

Google introduced Firefly as a software‑first clock synchronization design that separates rapid NIC‑to‑NIC consensus from UTC alignment. A distributed algorithm over a d‑regular random graph speeds convergence and scales as networks grow, while RTT filtering, path profiling, and optional Transparent Clock support reduce jitter and asymmetry. Tests showed consistent sub‑10ns NIC alignment and external sync well within 100µs thresholds for financial venues. Why it matters: precise time underpins fair ordering, telemetry fidelity, and deterministic fabrics for ML—delivered here on commodity data‑center gear instead of specialized timing appliances.

AI In The Console: Guardrails And Visibility

New generative interfaces in the AWS Management Console let operators query and visualize resource inventories and costs without manual aggregation. Amazon Q Developer artifacts present tables and charts with deep links, a relocated chat panel, and a Prompt Library to accelerate common queries. Outputs can expand to full screen for focused troubleshooting and cost analysis, with availability aligned to Regions offering the Developer tier.

Policy Autopilot brings open‑source IAM policy generation directly into agentic IDEs via the Kiro power, enabling one‑click setup and faster least‑privilege baselines during prototyping. The tool uses static analysis to propose starter permissions that teams review against production guardrails; the project remains open source for inspection and integration. Why it matters: embedding policy expertise where code is written reduces configuration drift and speeds safer defaults.

On offense, researchers summarized how adversaries fold generative AI into phishing, exploit development, and near‑autonomous campaigns. A report from CSOonline highlights AI‑crafted lures, deepfakes, alternative LLM platforms, theft of cloud credentials to run models, poisoning of agent memories, and attacks on AI infrastructure and supply chains. Recommended responses include tightened identity, least privilege, monitoring of AI/API usage and spend, and layered defenses tailored to agentic misuse.

Developer Supply Chains Targeted

Socket’s Threat Research Team exposed a worming campaign that typosquats npm packages and abuses compromised GitHub identities while reaching into AI coding assistants. According to The Hacker News, the operation (SANDWORM_MODE) steals SSH keys, AWS credentials, npm tokens, and environment secrets, injects rogue Model Context Protocol servers for assistants such as Claude Desktop, Cursor, VS Code Continue and Windsurf, and propagates by publishing infected packages and modifying repositories via API and SSH. Multi‑stage payloads use layered obfuscation; exfiltration chains span HTTPS to a Cloudflare Worker, uploads to attacker‑controlled GitHub repos, and DNS tunneling with DGA fallback. Vendors removed packages and infrastructure following reports. Why it matters: the blend of package abuse, CI poisoning, and AI‑assistant manipulation puts dev workstations and pipelines in the blast radius; teams should rotate credentials, audit repos and workflows, and inspect local assistant configurations.

Exploitation In The Wild

Active exploitation hit Ivanti EPMM via two zero‑days (CVE‑2026‑1281, CVE‑2026‑1340) that allow unauthenticated remote code execution. CSOonline reports automated scanning, rapid deployment of web shells and miners, and use of tools like the Nezha agent, with thousands of exposed instances observed. Ivanti issued emergency, version‑specific RPM patches for 12.x and advised restoring from known‑good backups for suspected compromises, with resets of passwords, service credentials, and certificates.

BleepingComputer relayed CISA’s addition of two Roundcube Webmail flaws to the KEV Catalog—CVE‑2025‑49113 (RCE) and CVE‑2025‑68461 (XSS). Agencies have three weeks to remediate under BOD 22‑01. Roundcube issued fixes in 2025, and historical exploitation of similar bugs by state and criminal actors underscores the persistent risk to internet‑facing email servers. Why it matters: KEV inclusion signals confirmed exploitation and prioritizes patching across federal networks.

Infosecurity covered an AWS‑attributed, financially motivated campaign that compromised more than 600 FortiGate devices across 55+ countries without exploiting a FortiGate‑specific vulnerability. The actor used multiple generative AI models to build tools for scanning, credential testing, VPN orchestration, and post‑VPN reconnaissance; attempts at domain compromise leveraged standard offensive tooling and often failed where targets were patched or ports closed. Recommended mitigations include credential hygiene, segmentation, post‑exploitation detection, and backup hardening.

The FBI warned of a sharp rise in ATM jackpotting, with 700+ incidents in 2025 costing banks over $20 million. Infosecurity reports malware like Ploutus abusing the XFS API after physical intrusion to issue hardware‑level dispense commands that bypass bank authorization. Suggested defenses pair physical controls with whitelisting, firmware integrity checks, strong logging, and EDR, plus staff awareness focused on fast cash‑out risks.