ACTIVE THREATS: CISA: Hackers now exploit max severity GitLab flaw in attacks • How a hole in Lenovos login system let hackers walk into 5,000 Dropbox accounts • The US military just turned off ad tracking on its phones. Maybe you should too • Hackers exploit Tencent app flaw to deploy GrayRabbit malware • CRPx0 ransomware: what you need to know ACTIVE THREATS: CISA: Hackers now exploit max severity GitLab flaw in attacks • How a hole in Lenovos login system let hackers walk into 5,000 Dropbox accounts • The US military just turned off ad tracking on its phones. Maybe you should too • Hackers exploit Tencent app flaw to deploy GrayRabbit malware • CRPx0 ransomware: what you need to know
The full HackWire archive — 3,892 stories, newest first.
Russian intelligence is conducting a phishing campaign using fake SMS messages impersonating Signal and WhatsApp support. The operation targets Ukrainian and Western government, military, and civil society officials to harvest credentials and sensitive information for espionage.
Education institutions faced 1,252 breaches in 2025, with 65% involving ransomware. The critical threat: vendor software vulnerabilities cascade across entire school ecosystems, compromising thousands of institutions simultaneously despite strong internal defenses.
Researchers found an attack targeting AI coding agents using clean GitHub repos. The exploit triggers error-recovery logic to gain shell access and steal API keys without any malicious code.
OpenAI released three GPT-5.6 tiers (Sol, Terra, Luna) balancing power, speed, and cost in a limited preview with U.S. government coordination on AI governance and cybersecurity safeguards.
236,000+ scam sites exploit DCloud's legitimate Uni-App framework for investment fraud, fake gambling, and phishing. Criminals sell pre-built templates in underground forums, enabling low-skill fraudsters to launch professional scams globally and steal tens of millions annually.
Trump's June 2026 executive orders set a 2030 deadline for post-quantum cryptography adoption across federal systems, compressing a predicted 10-year migration into five years. Organizations must identify quantum-vulnerable assets and patch legacy systems while managing astronomical, unbudgeted cost
AI transforms entry-level cybersecurity roles rather than eliminating them. Despite 44% of organizations reconsidering positions, 31% expect new roles to emerge, with change accelerating from years to months.
End-of-life open source software continues running unsupported in production, creating enterprise vulnerabilities. The Commonhaus Foundation launched an initiative to address this critical infrastructure security risk.
Cisco acquires Astrix and WideField to govern AI agents. As agent deployment surges (25% to 74% in 2 years), enterprises lack visibility over these privileged systems operating outside identity controls.
Autonomous AI penetration testing confidence plummeted 69% in one year (from 30% to 9%) as organizations discovered it cannot replace human security testing. The promised benefits of cost savings and 24/7 automation proved unrealistic.
FBI warns of targeted phishing attacks stealing Signal backup keys from government officials and journalists, exploiting recovery key management through social engineering to bypass encryption.
SharkLoader malware delivers Cobalt Strike Beacon in attacks targeting diplomatic and government agencies in Indonesia and Taiwan. The StrikeShark campaign marks an escalation in regional cyber espionage operations.
Russian intelligence is phishing Signal users for backup recovery keys—persistent credentials that remain valid even after password resets, enabling long-term message access. This escalates their March credential-theft campaign with a more sophisticated multi-stage attack on Signal's architecture.
Russian Intelligence Services are escalating a phishing campaign targeting messaging platforms (Slack, Teams, Zoom) to harvest credentials and establish persistence in government and critical infrastructure networks. Compromised accounts enable lateral movement, data exfiltration, and command-and-co
Chinese APT CL-STA-1062 deployed custom backdoor TinyRCT against Southeast Asian government and critical infrastructure since March 2022. Recent 2025 campaigns compromised 10+ organizations, signaling a shift to proprietary malware for persistent espionage access.
Attackers create fraudulent OpenAI organizations impersonating cybersecurity firms to trick employees and harvest corporate data. Dubbed "Poisoned Tenant," this campaign exploits OpenAI's legitimate infrastructure to bypass email security controls.
Polymarket's $3M theft came from a supply-chain attack injecting malicious code into the website, tricking users into approving fraudulent transactions. The company pledged full reimbursement.
Nebulock raised $25M Series A to enable proactive AI threat hunting as attackers use AI agents to compress breach timelines. The platform shifts defenders from reactive incident response to continuous threat detection.
State-sponsored actors breached an activist's iPhone using Cellebrite. A 630 GB supply chain theft exposed Apple and Tesla secrets. Five Eyes warns AI is compressing threat timelines.
Attackers breached Klue using compromised legacy credentials and stole OAuth tokens to access ~24 customers' Salesforce systems, exfiltrating business data. The attack went undetected for nearly a week until vendors disabled the integration on June 17.
Amazon Q Developer's CVE-2026-12957 flaw auto-executed malicious code in poisoned repos, silently stealing developers' credentials and API keys. AWS patched it in May after Wiz's April disclosure.
Amazon Q Developer (CVE-2026-12957, CVSS 8.5) allows code execution and AWS credential theft via malicious `.amazonq/mcp.json` files. Developers enable the attack by cloning a repository and trusting the workspace, exposing their full cloud environment and permissions to attackers.
AI agents now operate in enterprises with full access but no audit trails. Identity controls can't distinguish AI from humans, letting agents execute commands at machine speed without oversight.
**DirtyClone** (CVE-2026-43503) is a critical Linux kernel privilege escalation flaw allowing any local user to gain root access through memory corruption. With a public proof-of-concept exploit released June 25, 2026, it poses immediate risk to containerized environments and Linux infrastructure gl