CVE-2026-20253
Get tomorrow's brief in your inbox
Today: Anthropic disabled its two most advanced AI models after a US government export control order over jailbreak concerns. Microsoft patches failed to fix an on-prem SharePoint zero-day now under active attack. Chinese threat actors maintained 10-year persistence in an isolated network by backdooring PAM and OpenSSH authentication stacks.
Microsoft patches failed to fix on-prem SharePoint, which is now under zero-day attack
Microsoft's previous patches for on-premises SharePoint installations failed to remediate a vulnerability that is now being actively exploited in the wild. The zero-day allows attackers to compromise SharePoint servers that were thought to be patched. This affects on-premises SharePoint deployments only, SharePoint Online is not impacted.
Critical Splunk Enterprise Flaw Lets Attackers Run Code Without Authentication (CVE-2026-20253)
Splunk Enterprise versions below 10.2.4 and 10.0.7 contain a critical vulnerability (CVSS 9.8) that allows unauthenticated remote code execution through an exposed PostgreSQL sidecar service endpoint. The flaw exists because the endpoint lacks authentication controls, allowing network-reachable attackers to create or truncate arbitrary files, then weaponize database restoration to achieve RCE. watchTowr Labs published a working exploit chain that uses the /backup and /restore endpoints to dump attacker-controlled database contents, restore them with SQL execution, write arbitrary files using lo_export, and overwrite frequently-executed Python scripts with malicious payloads. EPSS score is currently low (0.001, 21st percentile), but public exploit details will drive opportunistic scanning.
Chinese hackers hijack auth flow, spy on isolated network for a decade
Chinese APT group Velvet Ant breached a large organization's isolated critical infrastructure network in 2016 and maintained undetected access for 10 years. The campaign, dubbed "Operation Highland" by Sygnia, began with internet-facing server compromises, then pivoted to an air-gapped network with no external connectivity. Velvet Ant modified Nginx configurations to proxy specially crafted HTTP requests through a compromised backend server to a FastCGI wrapper that established SSH connections into the isolated environment. Once inside, the attackers backdoored PAM modules (pam_unix.so) with nine distinct variants that accepted hardcoded passwords and harvested credentials. They also replaced OpenSSH components (ssh, sshd, scp) with trojanized versions that logged credentials and commands. By controlling the authentication stack, Velvet Ant gained visibility into all administrative activity and could bypass normal authentication. This extended the 2024 campaigns where Velvet Ant exploited F5 BIG-IP devices for three years and a Cisco NX-OS zero-day.
Ex-school district employee jailed for hacks on former employer
Ezekiel Dean Potter, 34, was sentenced to 21 months in prison for conducting a 21-month cyberattack campaign against his former employer, the Saydel Community School District in Iowa. Potter retained access credentials after his employment ended in April 2023 and repeatedly targeted the district's systems, deleting the Facebook page, removing employee access to educational platforms, and attempting password resets. He deleted the district's Apple School Manager account data (user accounts, passwords, device management), disabling MacBook and iPad management for a week. In January 2025, Potter accessed the Schoology learning management system through a Google administrator account, deleted an IT employee's account, and later deleted nine Gmail accounts including the IT director and superintendent. He began using a VPN after receiving Google security alerts. Federal investigators traced activity to IP addresses at Potter's subsequent employers. A USB drive recovered from Potter's desk at a former employer contained spreadsheets with Saydel account credentials. Potter pleaded guilty under the Computer Fraud and Abuse Act and must pay $59,668.81 in restitution.
NPM 12 Will Change Script Execution Behavior to Prevent Supply Chain Attacks
GitHub announced that NPM version 12, expected in July 2026, will block automatic execution of preinstall, install, and postinstall scripts from dependencies by default. This change responds to recent supply chain attacks (TeamPCP, Shai-Hulud worm) that abused automatic script execution during npm install to infect thousands of developers. The update also blocks native node-gyp builds unless explicitly allowed, and disables resolution of Git dependencies and remote HTTPS tarballs unless developers opt in. Starting with NPM 11.16.0, developers can run "npm approve-scripts --allow-scripts-pending" to generate an allowlist in package.json. Once committed, NPM 11.16.0+ will warn if install routines execute scripts not on the allowlist. The change also closes a code execution path where Git dependencies could override the Git executable via .npmrc.
US Gov asks Anthropic to ban 'foreign national' access to Fable, Mythos
The US government issued an export control directive on June 12 ordering Anthropic to immediately block access to its Fable 5 and Mythos 5 AI models by any foreign national, inside or outside the US, citing national security concerns. Because the scope includes foreign-born Anthropic employees, the company suspended access to both models for all users worldwide. Fable 5 had been rolled out free to Pro, Max, and Enterprise customers three days earlier on June 9. The directive stems from a reported jailbreak technique, which Anthropic says consists of asking the model to read a specific codebase and identify software flaws. Anthropic reviewed the demo and found only minor bugs that other publicly-available models (including OpenAI's GPT-5.5) can already discover without bypass. The company disputes that a narrow, non-universal jailbreak should trigger a recall affecting hundreds of millions of users, stating this standard would halt all new model deployments across the industry. All other Anthropic models, including Claude Opus 4.8, remain available. The UK's Minister for AI said the pause affects both US and UK customers.
Russians are posing as Signal support to launch phishing attacks
Russian threat actors are impersonating Signal support staff to conduct phishing campaigns. The attacks target Signal users through fake support communications designed to steal credentials or deliver malware. Signal does not provide direct user support through unsolicited messages, making these impersonation attempts a social engineering vector.
Google fires sueball at alleged Chinese phishers over AI-powered fraud ops
Google filed a lawsuit against a Chinese phishing operation called "Outsider Enterprise" that used Telegram-based infrastructure to send millions of scam texts impersonating trusted brands. The operation used AI to power fraud campaigns at scale. Google's legal action targets the operators and infrastructure supporting the phishing-as-a-service platform.
DEF CON Franklin project enlists hackers to harden critical infrastructure
DEF CON announced the Franklin project, expanding the successful voting village model to cover all critical infrastructure sectors across the entire conference. DEF CON founder Jeff Moss said the voting village reports have been so successful at identifying security gaps that the approach will now be applied conference-wide. The initiative brings hackers and critical infrastructure operators together to identify and remediate vulnerabilities before adversaries exploit them.
Ten years since the first corp ransomware, Mikko Hyppönen sees no end in sight
F-Secure's Mikko Hyppönen marked the 10-year anniversary of the first corporate ransomware attack with an assessment that the threat shows no signs of declining. Hyppönen noted that while the threat landscape continues to evolve, ransomware remains a stable career path for infosec professionals due to its persistence.
EQT buys majority share in Swiss cybersecurity biz Acronis
Private equity firm EQT acquired a majority stake in Swiss cybersecurity company Acronis at a valuation equivalent to $3.5 billion or higher for the entire firm. The specific portion sold was not disclosed. Acronis provides backup, disaster recovery, and cybersecurity solutions.
South Korea Hands Coupang a Record-Breaking $409 Million Data Privacy Fine
South Korea's data protection regulator issued a $409 million fine to e-commerce company Coupang, the largest privacy penalty in the country's history. The enforcement action continues South Korea's aggressive approach to data breach response, which has previously included suspending financial institutions' ability to enroll new customers following major data leaks. Details of the specific violations were not provided in the available coverage.
Today's stories highlight three persistent security challenges: authentication stack compromises enable decade-long persistence even in isolated networks, supply chain attacks continue to abuse default package manager behaviors, and AI model security is now subject to export controls. The Velvet Ant campaign demonstrates that air-gapped networks provide limited protection when attackers control the bridge between zones and can backdoor the authentication layer itself.