Here’s an overview of some of last week’s most interesting news, articles, interviews and videos:

Nono: Open-source sandbox for AI agents
AI coding agents run with the same permissions as their users, meaning they can access sensitive files, credentials, and production systems. A prompt injection, hallucinated command, or simple mistake can quickly turn that access into a security incident. To reduce the risk, Luke Hinds and Stephen Parkinson co-founded nolabs and released Nono, an open-source runtime that sandboxes AI agents at the operating system kernel, limiting what they can access and do.

Marathon Petroleum’s CISO on OT security automation, supply chain risk
In this interview with Help Net Security, Mary Rose Martinez, CISO at Marathon Petroleum, talks about what happens to security when automation reaches deep into refineries, pipelines, and terminals. She explains why the old idea of air-gapped operational technology has faded, how the Purdue model helps her team apply controls without stopping production, and where supply chain risk sits when vendors and their vendors hold the keys.

Shadow AI incident response begins with logs that may already be gone
In this Help Net Security interview, Brandy Wityak, VP of Complex Matters at LevelBlue, explains what happens in the hours after a shadow AI incident. She describes how quickly logs roll over, why firewall records of outbound traffic to AI platforms are often gone before responders arrive, and what regulators look for when they assess whether a company did enough.

Your AI agents can reach data no one approved
A credential expired. An AI agent kept using it anyway, and a mid-sized company’s systems went down for a quarter’s worth of trouble before anyone traced the failure back to a non-human account no one had been logging.

The energy sector’s OT cybersecurity talent is retiring faster than it can be replaced
A ransomware hit lands a chemical plant in a safe state. Nobody is hurt, the site holds steady, and the operators begin the restart. The systems stay down. Every attempt to bring them online meets encrypted processes and altered configurations. The outage runs into weeks, and the losses travel down the supply chain in both directions. Refineries, chemical plants and pipeline operators carry this exposure across control systems that run 20 to 40 years.

An AI agent can pass every safety check and still leak secrets
A pull request lands with a tidy bug report in the description. A bot reads it before any person does, pulls a few shell commands out of it, gets them approved, and posts the output back on the thread. The maintainer reads the whole exchange the next morning. Elad Meged, a founding engineer at Novee Security, ran that sequence against three vendors’ own repositories, in the configurations those vendors ship by default.

Top companies to visit at Black Hat USA 2026
Black Hat USA 2026 returns to Mandalay Bay with a re-engineered six-day program designed to spark innovation, challenge assumptions, and unite the global security community. From cutting-edge innovators to industry veterans launching new offerings to rising stars shaking up the status quo, these exhibitors are bringing something special to the floor this year. Make time in your schedule to stop by, because your next big opportunity might be waiting.

200 new CVEs a day and no realistic way to patch them all
Ryan Dewhurst, CEO at KEVIntel, explains how his team confirms exploitation that CISA’s catalog has not listed yet. He describes a global honeypot sensor network, AI triage, and human verification in a lab before a vulnerability reaches the public feed.

Data breach cost 2026 averaged $4.99 million, AI attacks ran higher
More than one in four organizations hit by a malicious attack over the past year say AI drove it. Those breaches averaged about $1 million above the malicious attacks that ran without AI.

Aviation cyber risk sits on the ground, the blindness sits in the air
In this interview with Help Net Security, Eliran Almog, CEO of Cyviation, explains why airline cyber losses happen on the ground while the aircraft stays unmonitored. He walks through GNSS jamming that leaves no trace in a SIEM, and a PX4 Autopilot flaw his team disclosed where drone command channels accept unsigned messages.

PoC exploit released for critical AD CS domain-takeover flaw (CVE-2026-54121)
Security researchers who discovered and reported CVE-2026-54121 (aka “Certighost”), a critical privilege elevation vulnerability in Active Directory Certificate Services (AD CS), have released a proof-of-concept (PoC) exploit for and technical details related to the flaw.

JetBrains fixes critical unauthenticated RCE in TeamCity On-Premises (CVE-2026-63077)
JetBrains has fixed a critical vulnerability (CVE-2026-63077) affecting TeamCity On-Premises and is urging admins to upgrade self-hosted servers as soon as possible.

Hugging Face breach reignites open-weights debate, raises liability questions
The first publicly documented cyberattack run end-to-end by an autonomous AI was an OpenAI benchmark test that escaped its sandbox and breached Hugging Face. In an incident post-mortem compiled with the input from Hugging Face and several hundred members of Cloud Security Alliance’s CISO community, the nonprofit organization laid out the most salient points for security leaders and advised on what they should do next.

Cisco FMC static credentials exploited by attackers (CVE-2026-20316)
A static credentials vulnerability (CVE-2026-20316) in Cisco Secure Firewall Management Center (FMC), a platform for centrally managing multiple Cisco Secure Firewall devices across a network, is being leveraged by attackers, CISA warned.

Laundry Bear’s new Microsoft Exchange attack triggers on email open (CVE-2026-42897)
Russia-affiliated cyber espionage group Laundry Bear (aka Void Blizzard, aka TA488) is exploiting CVE-2026-42897, a cross-site scripting vulnerability in Microsoft Exchange, to target US and European government entities and a variety of private sector organizations via email. The warning comes from Proofpoint, who detected emails carrying the concealed exploit hitting inboxes.

What the identity attack surface looks like when trust becomes the target
In this Help Net Security video, Joel Moses, VP, Strategic Engineering at F5, explains how attackers use identity instead of breaking through it. He walks through MFA fatigue, session token theft, and consent given to malicious applications, using the 2022 Uber breach as an example.

Impersonation protection: How to protect your executives when the truth isn’t clear
How do you protect your executives when truth doesn’t seem to be truth anymore? It’s a question BlackCloak Founder and CEO Dr. Chris Pierson recently discussed with SVP of Product Matt Covington.

GitHub delays version updates so malware gets caught first
Automated dependency update tools normally open pull requests as soon as a new package version is released, but that speed can backfire. In September 2025, attackers published malicious versions of popular npm packages such as chalk and debug, and although they were removed within about two hours, that was enough time for update bots to propose them to downstream projects. To reduce this risk, GitHub introduced Dependabot cooldown, which delays non-security update pull requests for at least three days by default.

Google changes how it names cyber threat actors
Google Threat Intelligence Group (GTIG) has started using a new naming system for the threat actors it tracks. The change comes after Mandiant and Google’s Threat Analysis Group (TAG) merged into one unit, leaving the company with two separate naming schemes built up over years.

Tech giants form alliance to put open AI in cyber defenders’ hands
NVIDIA and a group of tech companies have formed an alliance to promote the use of open AI models in cybersecurity, days after OpenAI disclosed that one of its own AI models breached Hugging Face’s systems during an internal security evaluation. The new group, called the Open Secure AI Alliance, builds on work already underway at the Linux Foundation’s Akrites initiative and the Open Source Security Foundation (OpenSSF).

Microsoft unveils MAI-Cyber-1-Flash, promises cybersecurity AI at half the cost
Microsoft has introduced MAI-Cyber-1-Flash, a security-focused AI model built into MDASH, the company’s multi-agent vulnerability identification and remediation system.

Call of Duty Mobile scam uses fake free points giveaway to hijack players’ accounts
Call of Duty Mobile players should watch out for a phishing campaign disguised as a free Call of Duty Points giveaway, Malwarebytes researchers have warned. Victims are asked to log in with their email address and password to claim free Call of Duty Points (CP), the game’s premium currency, before being redirected to a second page requesting their 2FA code.

AI took more than junior developer jobs and the bill comes later
A ticket comes in for a small bug fix. Hand it to the junior on your team and you wait a day, review something that half works, and sit down to explain what went wrong. Describe it to Claude and the patch merges before lunch.

Coca-Cola confirms hackers stole data in Fairlife ransomware attack
Coca-Cola has confirmed that the ransomware attack on its dairy subsidiary Fairlife involved the theft of company data, weeks after the incident temporarily halted production at its US facilities.

VERITAS project could change the way scientists secure AI
The AI models, datasets, and automated systems researchers depend on can be compromised in ways conventional cybersecurity tools aren’t designed to detect. A new project called VERITAS (VERified Infrastructure for Trustworthy AI in Science) aims to close that gap by establishing AI Assurance as a core function of scientific research infrastructure.

Exposed BMCs hand out password hashes before login
An attacker who reaches UDP port 623 on a server’s baseboard management controller can ask it for a password hash and receive one before logging in. The exchange is part of the IPMI 2.0 handshake, built on an authentication protocol introduced in 2004.

WhatsApp brings end-to-end encrypted voice and video calls to the web
WhatsApp has launched support for voice and video calls on the web, allowing users to make and receive calls directly from their browser without installing the desktop app.

Stolen Meta and Google ad accounts are worth more than the money they hold
Ad account theft, the systematic hijacking of Meta Business Manager and Google Ads accounts, has grown into a commodity-driven cybercrime economy complete with tiered pricing, escrow services, and money-back warranties for stolen accounts. Public reporting on this topic tends to focus on drained ad budgets, but according to Mimecast, that’s often a short-lived gain for attackers.

Cloudflare reveals what’s behind major internet outages
Storms, earthquakes, and infrastructure failures disrupted internet access throughout the second quarter, while governments deliberately shut networks down, according to Cloudflare’s latest Internet Disruption Summary.

Tengu botnet reboots Linux devices to survive removal
A new Mirai-derived IoT botnet can force an infected Linux device to reboot once its main process is killed, giving its persistence mechanisms another opportunity to relaunch it, Nozomi Networks Labs has found.

Coordinated cyberattack hits more than 30 Minnesota water utilities
A coordinated cyberattack on July 26 and 27 hit operational technology (OT) systems at more than 30 community water utilities across Minnesota, prompting an immediate response from Minnesota IT Services (MNIT) to contain the threat.

Attackers are using Microsoft’s legitimate login system to camouflage phishing attacks
Attackers are moving away from fake Microsoft login pages in favor of abusing Microsoft’s own authentication system, letting phishing campaigns slip past the warning signs employees are trained to spot, according to Check Point.

AI takes on a bigger role in finding Chrome vulnerabilities
Google has expanded the use of AI in Chrome’s security workflow, using it to find vulnerabilities, triage bug reports, generate patches, and review code to shorten the time between discovering software flaws and delivering security updates.

Anthropic’s Claude breached three companies during security tests
Anthropic has disclosed that its AI model Claude gained unauthorized access to the systems of three different organizations during cybersecurity evaluations. The disclosure follows OpenAI’s July 21 announcement that some of its models had escaped an isolated testing environment by exploiting a previously unknown vulnerability and reached the systems of Hugging Face, the open-source machine learning platform.

Criminals used AI and children’s coding software to build a multimillion-dollar ad fraud empire
A security investigation into inexpensive Android TV boxes led researchers to an ad fraud operation that had remained unnoticed for several years. According to Bitsight, the operation, named Fuyao, uses preinstalled Android apps, device identity spoofing, AI-generated websites, and residential proxy services to generate advertising revenue without device owners’ knowledge.

Claude Opus 5 sharpens coding and cybersecurity work on AWS
Claude Opus 5 went live on Amazon Bedrock and Claude Platform on AWS. Anthropic says the model improves on Claude Opus 4.8’s cyber capabilities, coding through cybersecurity. Anyone with an AWS account in a supported region can call it. On higher-risk requests, Opus 5 hands the job back to Opus 4.8, the older model. The user sees a notice when that happens. API customers can configure the fallback.

Product showcase: LastPass Authenticator brings Face ID, Apple Watch, and cloud backup to 2FA
LastPass Authenticator is a free app that provides two-factor authentication (2FA) for accounts and any service that supports time-based one-time passwords (TOTP). It supports push notifications for one-tap approvals and generates six-digit verification codes for online accounts. The app is available for iPhone, iPad, Apple Watch, and Android devices.

AWS gives DevOps teams an AI investigator for firewall incidents
AWS DevOps Agent helps administrators inspect logs, review firewall rules and network paths, identify configuration changes that caused AWS Network Firewall to block traffic, and restore connectivity. The service is an AI-powered operations assistant for DevOps and SRE teams that investigates and troubleshoots application and infrastructure issues.

ChatGPT joins the most impersonated brands in phishing attacks
Microsoft continued to be the most impersonated brand in Q2 2026, accounting for 23% of all brand phishing attempts. LinkedIn, Google, Apple, and Amazon followed, with the five brands together making up more than half of all brand phishing attempts tracked during the quarter, according to Check Point’s Q2 2026 Brand Phishing Report.

AWS to retire Shield Advanced L7 automatic mitigation on January 1, 2027
AWS Shield Advanced, a managed service that protects applications from external threats, is adding the Anti-DDoS managed rule group, designed for application-layer (L7) DDoS protection, to eligible web access control lists (ACLs) in Count mode.

Android malware detection collapses when the context stage comes out
A phone backup app requests storage, contacts, SMS, and call logs. A device management app requests even more. Run either through an Android malware detector, and it may be flagged as malicious. Researchers at Singapore Management University and Nankai University found that six widely used Android malware detectors.

Specter: Open-source NFC reader bug sweep for Flipper Zero
Specter is a Flipper Zero app that finds powered NFC readers by listening for the radio field they give off. The readers it hunts work at 13.56 MHz. The onboard ST25R3916 carries a hardware external-field detector, the same circuit that lets the device emulate a card and register when a reader starts talking to it. Specter reads that one bit, hundreds of times a second, with its own transmitter dark.

Exposed credentials are giving attackers a head start many organizations don’t see
Compromised credentials can remain active long after passwords are created, leaving organizations trying to identify exposed accounts before attackers can use them. The 2026 Credential Risk Report from Enzoic shows growing awareness of the problem, but monitoring and response capabilities continue to lag.

Product showcase: Dashlane Password Manager is more security toolkit than password vault
Dashlane is a password manager for individuals and families that stores passwords, passkeys, payment cards, personal information and secure notes in an encrypted vault. It also includes a password generator, password health reports, an authenticator, credential sharing, dark web monitoring and phishing protection.

CISA sets a new SBOM baseline
The US Cybersecurity and Infrastructure Security Agency (CISA), together with its co-authoring partners, has released the 2026 Minimum Elements for a Software Bill of Materials (SBOM), replacing the 2021 guidance published by the National Telecommunications and Information Administration (NTIA).

AI agents are changing where cybersecurity seed funding lands
Founders pitching a cybersecurity seed round this summer are joining a line that keeps getting longer. Product Hunt launches hit their highest level since late 2023 last quarter, and the Census Bureau’s count of high-propensity business applications kept climbing. Seed deal volume in cyber ticked down. Those figures come from the Q2 2026 Insights report published by DataTribe, an early-stage cybersecurity investor.

Companies push AI, sysadmins keep it on a short leash
In 2024, sysadmins expected AI to automate patch management optimization, vulnerability prioritization, infrastructure monitoring, and incident response within two years. Action1’s 2026 Survey Report: AI Impact on Sysadmins found that those expectations proved overly optimistic.

Cybercrime goes subscription: AI, malware and infrastructure on demand
Cybercrime has become a commercialized ecosystem where criminals can buy or rent nearly every capability needed to launch sophisticated attacks. These services provide anonymity, plausible deniability, and access to short-lived infrastructure that is difficult to detect, attribute, and disrupt, enabling low-skilled actors to operate at scale, according to the Infoblox 2026 Threat Landscape Report.

Download: The High-Performance Team Playbook
Get practical insight from teams who’ve built, scaled and handed over engineering functions at enterprise level.

ShutterGap: Aryon Security finds 3.7M AWS cloud resources exposed beyond CSPM/CNAPP visibility
Research from Aryon reveals that each year, 3,731,699 short-lived cloud resources containing highly sensitive information are publicly exposed. This impacts any organization using AWS services that support public sharing.

Cybersecurity jobs available right now: July 28, 2026
We’ve scoured the market to bring you a selection of roles that span various skill levels within the cybersecurity field. Check out this weekly selection of cybersecurity jobs available right now.

New infosec products of the week: July 31, 2026
Here’s a look at the most interesting products from the past week, featuring releases from BlackCloak, Contrast Security, Dropzone AI, PortSwigger, Realm Security, Reco, Root Evidence, and ZeroFox.



Source link