Week in review: Salesforce and ServiceNow portals exposed for 17 months, exploited Metabase 0

Week in review: Salesforce and ServiceNow portals exposed for 17 months, exploited Metabase 0-day Here’s an overview of some of last week’s most interesting news, articles, interviews and videos: GitHub Dependabot malware alerts now cover eight ecosystems GitHub has flagged npm malware since March 2026. Anyone pulling in a bad PyPI, Maven, RubyGems, NuGet, Go, crates.io, or PHP Composer package has had no such warning, because GitHub’s malware detection only ever watched one ecosystem. That changed this month. Dependabot malware alerts, which had run on npm data alone, now draw on all eight ecosystems the moment a user turns malware alerts on. Dependabot itself already runs across more than 30 million repositories and 34-plus package ecosystems overall, which gives a sense of the scale the malware pipeline now has to operate at. An AI tool found 84 flaws in 5G network software and 23 of them still have no fix Researchers at Nanyang Technological University turned a set of AI agents loose on the software that runs 4G and 5G phone networks, and the agents came back with 84 security flaws nobody had reported before. Developers have confirmed 83 of them, and 81 now carry CVE numbers. Post-quantum migration gets harder when every user holds a key In this Help Net Security interview, Christopher Smith, CEO of Quantus, discusses what cryptographic inventories turn up in banks and hospitals, including default passwords and admin keys still held by former employees. A stranger has been reading Salesforce and ServiceNow portals worldwide for 17 months Most security stories start with something broken. This one starts with everything working as designed. Researchers at Reco have been tracking a campaign they call City-Forum, named after a domain registered in 2002, abandoned, and now resolving to a generic rented server from a German hosting provider. From that server, someone has been pulling records out of Salesforce and ServiceNow portals around the world. N-able ships second N-central hotfix as attackers keep exploiting CVE-2026-18577 To help customers fend off ongoing attacks, N-able released a second security hotfix for N‑central, its monitoring and management (RMM) solution popular with managed service providers (MSPs). Metabase zero-day exploited to access Framework customer data Framework, the San Francisco-based company that designs repairable and upgradeable laptops, has suffered a data breach after attackers managed to exploit a zero-day vulnerability in the Metabase business intelligence service. According to the notification sent to affected Framework customers, the attackers accessed names, email addresses, phone numbers, physical addresses, and login IP addresses, but not payment information or records related to orders. Microsoft patches 400+ vulnerabilities, one zero-day under attack (CVE-2026-68820) Microsoft’s August 2026 Patch Tuesday delivered security fixes for 400+ vulnerabilities, including one that has been exploited in zero-day attacks (CVE-2026-68820) and three that were publicly disclosed prior to the release of the patches. Cisco fixes vulnerability exploited to DoS its firewalls (CVE-2026-20349) A high-severity vulnerability (CVE-2026-20349) is being leveraged by attackers to temporarily interrupt the operation of Cisco firewalls, the company has confirmed. The flaw has been added to CISA’s Known Exploited Vulnerabilities catalog and needs to be remediated by US civilian federal agencies by August 14, 2026. Who will be the Stanislav Petrov in your organization? Recent reports of “rogue AI” systems hacking companies reminded Brian Honan, CEO, BH Consulting, of Stanislav Petrov. In 1983, Soviet computers falsely detected a US nuclear missile launch. Instead of trusting the system, Petrov applied human judgement and correctly identified it as a false alarm, potentially preventing nuclear war. That is why the recent incidents involving OpenAI and Hugging Face, Anthropic, Meta, and the UK’s AI Security Institute deserve the attention of CISOs and boards. 338 million attack simulations reveal the state of enterprise defense First, a bit of good news: Enterprise defenses are recovering. However, it’s a narrow recovery, with a twist. Today, organizations are better at stopping loud attacks but have barely moved the needle at all against the quiet ones. This data, and a lot more, comes straight from the newly published Blue Report 2026, the fourth annual comprehensive study from Picus Labs. Four corporate investigation mistakes organizations make under pressure In this Help Net Security video, Christine Gadsby, VP and Chief Security Advisor at BlackBerry, explains why corporate investigations go wrong before the forensic team arrives. The first hours matter more than leaders assume. Access gets granted, conversations start, and decisions get made that later affect chain of custody, privilege, and how regulators judge the process. How to report an AI Act violation in the EU The EU’s fight to regulate AI models entered a new chapter on 2 August 2026, when the European Commission’s AI Office and national authorities began enforcing the AI Act. It creates a common set of rules for AI systems used or sold in the EU, with the goal of encouraging innovation while protecting people’s safety and fundamental rights. Chainloop: Open-source evidence store and policy engine for the software supply chain Chainloop is an open source evidence store for the software supply chain. A command line tool runs inside a GitHub Actions, GitLab, Jenkins, or Dagger pipeline, picks up what the build produced, uploads those files to content-addressable storage, and references each one in a signed in-toto attestation. in-toto is a specification for recording who ran which step of a build, so the record can be checked afterward. Microsoft Entra ID is removing an extra MFA hurdle for Windows Hello and macOS PSSO users Microsoft is changing how Entra ID handles MFA for people who sign in with Windows Hello for Business (WHfB) or macOS Platform Single Sign-On (PSSO). The rollout reaches worldwide and GCC tenants starting early October 2026, with completion expected by late November. Cyberattack on Steam hardware shipper leaks names, addresses, and order data Video game publisher Valve is alerting customers in Europe to a data breach at CEVA Logistics, its Steam hardware shipping partner. Reports from affected customers began surfacing on social media earlier today, after Valve started sending out data breach notification emails. Previously unseen entry vector used to breach Polish energy plant The December 29 cyberattack on a Polish combined heat and power (CHP) plant was the first observed case of attackers gaining access to an OT network through a private APN, according to CERT Polska. The private APN is a dedicated mobile network that a Distribution System Operator (DSO), the company running the local electricity grid, sets up with a mobile carrier. GPT-5.6-Cyber refuses security researchers’ requests far less often GPT-5.6-Cyber is a new OpenAI model built on GPT-5.6 Sol, trained to find zero-day vulnerabilities and build exploit chains, with fewer refusals on higher-risk, dual-use work. Model is available only through Daybreak Red, the higher tier of OpenAI’s vetted access program for cybersecurity professionals. Ransomware gangs don’t need control system access to disrupt industrial production Disrupting IT systems that support industrial environments can be enough to interrupt production, even when ransomware operators do not gain direct access to industrial control systems (ICS), according to Dragos. The company identified 1,140 ransomware incidents involving industrial organizations in the second quarter of 2026, up 12% from 1,020 in Q1. Malicious SIMs can hijack smartphones, steal files, and lock them onto 2G Researchers have found that compromised or malicious SIM cards can issue commands to some smartphones and cellular-connected devices, allowing attackers to steal information, disrupt communications, downgrade connections to 2G, and in some cases execute code. Ready-made $500 kit puts a crypto scam within anyone’s reach A seller on a cybercrime forum is offering a ready-made scam kit for $500, complete with an admin panel that tracks victims, checks their crypto wallets for value, and inflates fake balances to squeeze out more money, Malwarebytes found. Split-second deepfake glitch blows digital certificate fraudster’s cover Spanish police have arrested a man in Murcia accused of using deepfake software to trick a certificate provider’s video identity checks in an attempt to obtain digital signatures he could use for financial fraud. Lazarus hackers pair fake job offers with Windows zero-day exploit The North Korea-linked Lazarus group is using fake job offers, trojanized PDF software and a Windows zero-day in attacks aimed primarily at the defense sector, Check Point researchers have found. Signal’s new security feature checks if your encrypted chats were tampered with Signal has introduced a feature called automatic key verification, giving users a new way to confirm that nobody has secretly interfered with their encrypted chats. 153GB of stolen credentials surface after LiteLLM supply chain attack A massive 153GB archive stolen during the LiteLLM supply chain attack exposes credentials and other sensitive data linked to thousands of corporate domains, including AWS, Samsung, Cisco, and Salesforce. Hudson Rock says it obtained and analyzed the archive, which contains 433,909 files, and attributed 118,829 CI runner dumps to 2,488 corporate domains. Attackers exploit critical SharePoint flaw after PoC goes public (CVE-2026-55040) Threat actors have begun exploiting a critical Microsoft SharePoint flaw following the release of proof-of-concept (PoC) exploit code by Rapid7. Tracked as CVE-2026-55040, the vulnerability was patched by Microsoft as part of its July 2026 Patch Tuesday updates. White House authorizes private US companies to hack foreign criminal networks President Trump signed a National Security Presidential Memorandum on August 12 allowing vetted private companies to run offensive cyber operations against foreign threat actors, under the control and oversight of the US government. Ukrainian police raid 94 fraudulent call centers, seize $2 million Ukrainian police have disrupted 94 fraudulent call centers during a nationwide operation that involved more than 400 searches and the seizure of thousands of computers, phones, and SIM cards. New Android malware relays bank cards to fraudsters while victims still hold them Group-IB researchers discovered WindRelay, a new Android malware built to capture live payment card data over NFC (Near Field Communication) and relay it to attackers in real time. WindRelay is paired with the SpyNote remote access trojan, which gives attackers remote access to a victim’s device. 71% of CISOs spend 10+ hours on board reports Boards want evidence that security controls and architecture reduce business risk, expressed in terms of resilience, consequence, and decision relevance. Translating technical findings into business language remains a major time burden for CISOs, who are calling for simpler data delivery, better frameworks, and better context. Pulse Security AI’s The CISO-Board Communication Gap report found that board members bring external information into discussions while many organizations still lack a formally defined cyber risk appetite. Product showcase: Enpass Password Manager breaks away from the proprietary cloud model Enpass is a password manager that stores passwords, passkeys, payment cards, identities, secure notes, software licenses, and other sensitive information in encrypted vaults. Vaults remain on the device or in a cloud storage service selected by the user. OpenAI locks down Astra over potential critical cyber capabilities OpenAI’s internal evaluation of its upcoming model, Astra, found significant advances in agentic coding and cybersecurity, leading the company to conclude that it cannot rule out the model reaching the critical capability level for cybersecurity under its Preparedness Framework. Anthropic to put AI in charge of reviewing Claude Code actions by default Anthropic will make auto mode in Claude Code the default for new sessions on Pro, Max, and Team plans starting August 14. Users who previously selected a different default may receive a one-time prompt asking whether they want to switch to auto mode. Your security vendor gets the frontier cyber model, you get the findings Selected red team specialists can now use OpenAI’s cyber models to find and exploit weaknesses in client applications and infrastructure. Those clients never get the models themselves. That split is the design of the Daybreak Cyber Partner Program, which OpenAI expanded on August 10: access to the underlying models stays with the approved partner and is not transferred directly to the customer. Locking your ssh-agent exposed local-only keys until OpenSSH 10.5 Lock your ssh-agent and it should sit there refusing to sign anything until you unlock it. In OpenSSH 10.4, locking it also switched off the check that tells the agent whether a request came from your own machine or arrived down a forwarded connection from a remote server. The fix shipped today in OpenSSH 10.5. AI deployments are stretching enterprise security to its limits CISOs and CTOs expect AI deployments to increase their organizations’ attack surface by an average of 14% over the next year. Nearly all lack visibility into AI deployments, and 90% are concerned about employees using unapproved AI tools outside formal oversight, according to NetFoundry’s 2026 State of Secure AI Access survey. PentestGPT: Open-source automated penetration testing agentic framework PentestGPT is an open-source penetration testing agent that points a large language model at a target and lets it work. In its default mode it runs recon, then exploit, then walkthrough, each stage feeding the next. Switch it to pentest mode and the stages become asset discovery, vulnerability identification, report. No human sits in the loop. Chrome’s anti-abuse protections block 7 billion unwanted Android notifications daily Google Chrome’s latest measures against abusive web push notifications include automatically revoking notification permissions for inactive and suspicious websites, helping reduce scams, phishing attempts, and other deceptive content. Wireshark 4.6.8 patches 28 security bugs, nine in file parsers Wireshark 4.6.8 fixes 28 security bugs in the protocol analyzer, and nine of them fire when someone opens a saved capture file. Those nine sit in file parsers, the code that reads a capture off disk before any dissection begins: pcapng, Endace ERF, Tektronix K12xx, BUSMASTER, Catapult DCT2000, Gammu DCT3, 3gpp phone logs, TTX Logger, and, on Windows only, Ixia IxVeriWave and Vector Informatik BLF. An attacker never has to touch your network for these. They only have to get you the file. Product showcase: Is this image real? Slop or Not investigates Slop or Not is an AI text and image detector for iPhone and Mac that runs entirely offline, with no account required. It uses on-device AI models powered by the Apple Neural Engine to detect AI-generated content. DDoS attacks hit record scale as 1 Tbps+ campaigns become more common DDoS attacks grew in scale during the first half of 2026, bringing larger traffic floods, shorter attack durations, and increasingly automated campaigns. Cloudflare’s H1 2026 DDoS Threat Report shows threat actors relying on multi-vector techniques and large-scale network-layer attacks to disrupt online services across multiple industries. 17 draft Cyber Resilience Act standards are open for comment A company selling a connected toy in Europe must show by the end of 2027 that the product meets the Cyber Resilience Act. The law states what manufacturers have to achieve and stops there, which leaves the toymaker to work out the technical detail alone. Seventeen draft standards, now open for comment, supply that detail. Weak IAM affects up to 98% of cloud environments Misconfiguration remains one of the leading threats to cloud environments because a single configuration error can result in public network access, unrotated keys, missing encryption, exposed services, and logging gaps. CISA now mandates baseline cloud configuration practices for US federal agencies. The hardest part of agentic AI may be rebuilding the business Organizations expect AI agents to change how work gets done, driving productivity and growth while allowing employees to focus on higher-value tasks. Few, however, have the processes and workflows needed to realize those benefits, according to Deloitte’s latest research. AWS Certificate Manager sets 2027 end date for email-validated certificate renewals AWS Certificate Manager (ACM) will phase out email validation for public certificates throughout 2027, ahead of the Certification Authority/Browser (CA/B) Forum’s March 15, 2028 deadline for ending email-based domain validation. OpenAI’s GPT-5.6 Sol runs up to 14× faster with Ultrafast mode OpenAI’s GPT-5.6 Sol on Ultrafast mode is available in limited preview to a select group of customers, launching first through the OpenAI API. The company says the service runs up to 14 times faster than Standard processing and generates up to 750 output tokens per second. Ultrafast is powered by Cerebras as part of the companies’ partnership on ultra-low-latency inference. Cybersecurity jobs available right now: August 11, 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: August 14, 2026 Here’s a look at the most interesting products from the past week, featuring releases from A10 Networks, ScienceLogic, Searchlight Cyber, and SelectHub.