Security
Berlin Ransomware Breach Exposes 1.44M Files After Seven-Day Isolation Gap
Rhysida stole 1.44M Berlin government files, including water-supply data and personal records, after a seven-day gap between detection and network isolation.
This edition was produced with artificial intelligence. Text and voice are generated automatically.
Berlin Ransomware Breach Exposes 1.44M Files After Seven-Day Isolation Gap
Berlin’s state government confirmed on August 31, 2026, that the Rhysida ransomware group stole data from its administrative network during a five-day window in August. Forensic investigators traced the breach to a seven-day gap between the city’s first detection of the intrusion on August 7 and its disconnection of compromised departments from the central Berliner Landesnetz network on August 14. The gap allowed the group access to 1.44 million government files, including water-supply vulnerability assessments, plaintext credential files, and personal records of 12,076 Berliners.
Berlin’s Seven-Day Ransomware Isolation Gap Let Rhysida Steal Critical Infrastructure Data →
Berlin Refuses Ransom, Faces Data Release Deadline and Regulatory Scrutiny
Governing Mayor Kai Wegner said at an emergency Senate session that Berlin would not pay the ransom demand of 30 bitcoin, roughly €2 million. “The state of Berlin is being blackmailed,” he said. The auction deadline for Berlin’s data is expected to fall around September 4, 2026. If Rhysida releases any portion of the claimed 5.79 TB after the deadline, Berlin will face a second wave of GDPR Article 34 exposure and citizen notification obligations. Regulators may determine that the relevant awareness date for the 72-hour notification window was August 7, placing the deadline on August 10.
Berlin’s Seven-Day Ransomware Isolation Gap Let Rhysida Steal Critical Infrastructure Data →
Berlin Breach Details: Rhysida’s Tactics, Data Categories, and Broader Impact
Rhysida’s leak-site listing breaks down into eleven file categories, including 124,823 mapping and geodata files, 5,941 files described as containing passwords, and 8,110 infrastructure files including vulnerability assessments of Berlin’s water supply. The listing claims records on 12,076 named individuals, 16,389 email addresses, 11,963 phone numbers, and 148 IBANs. From August 14 to August 23, departmental staff lost access to the internet, external email, and shared district databases, and housing-benefit applications and payments were suspended for more than 50,000 households during that period.
Berlin’s Seven-Day Ransomware Isolation Gap Let Rhysida Steal Critical Infrastructure Data →
Berlin Breach Raises NIS2 and GDPR Compliance Questions
Berlin’s departments may qualify as essential entities under NIS2’s critical-infrastructure provisions. The NIS2 Directive imposes a three-stage incident-reporting structure: a 24-hour early warning, a 72-hour detailed notification, and a one-month final report. The European Commission’s NIS2 amendment proposal (COM(2026) 13 final, January 2026) is the first EU legislative instrument to address ransomware payments in a structured way, requiring entities to disclose on regulatory request whether a ransom was paid and to whom. Germany’s BSI, actively involved in the Berlin investigation, has consistently advised against ransom payments.
Berlin’s Seven-Day Ransomware Isolation Gap Let Rhysida Steal Critical Infrastructure Data →
US and CrowdStrike Dismantle Sality Botnet After Two Decades
U.S. law enforcement officials and cybersecurity company CrowdStrike announced on Tuesday the dismantling of “Sality,” a Russian hacking operation active for two decades. U.S. officials said they seized the web domains hackers used to take over computers for spam distribution, distributed denial-of-service operations, and cryptocurrency theft. CrowdStrike said it cut off a network of compromised computers from the mastermind controlling it. CrowdStrike began dismantling the botnet on Monday at its Day Zero threat intelligence summit in Las Vegas.
Sality Takedown: How CrowdStrike Turned the Botnet’s Own Architecture Against It
Sality’s peer-to-peer architecture allowed it to receive commands through a diffuse network of compromised machines, making it particularly resistant to law enforcement action. CrowdStrike said it turned that strength against Sality by seeding the network with bogus information that tricked the botnet’s components into cutting themselves off from their creator. CrowdStrike researcher Tillmann Werner said reverse-engineering the botnet’s structure, finding weak points, and building the infrastructure needed to knock it down required painstaking work. “This was the most complex botnet takeover we have ever done,” Werner told Reuters.
Sality Botnet Dismantled: What’s Next for the Creator and Victims
David Watson, director of nonprofit security group The Shadowserver Foundation, which was also involved in the takedown, said Sality was “quite old-school” but could still be dangerous. “It’s still a vector into a lot of organizations,” Watson said. He said the next step would be to see what Sality’s creator, who has yet to be publicly identified, did to regain control of or re-create the botnet. “What does he do?” Watson said. “Does he fight back?”
Dropbox Breach Traced to Lenovo SSO Authentication Flaw
Multiple Dropbox users were emailed by the company advising them that unauthorized access had occurred to their accounts. The breach took place between August 4 and August 21, 2026, and Dropbox logs showed no evidence that files were viewed or downloaded. The incident stemmed from a problem with a single sign-on (SSO) option using Lenovo IDs. Dropbox’s investigation determined that an issue with Lenovo’s email verification process allowed an unauthorized party to register a Lenovo ID using a victim’s email address and then use that ID to log into the associated Dropbox account.
Dropbox breach seemingly caused by egregious authentication failure →
Dropbox SSO Flaw: Attack Chain and Fix
The attack chain involved the attacker compiling target email addresses, registering a Lenovo ID as the victim without needing inbox access because Lenovo’s verification step was missing or bypassable, then using federated sign-in so Lenovo’s authorization server issued a token matching the victim’s email claim. Dropbox then resolved the email claim to the existing account and minted a session without a password prompt, step-up authentication, or consent to link the new identity. Dropbox has since fixed the flaw and expired all sessions previously authenticated through a Lenovo ID.
Dropbox breach seemingly caused by egregious authentication failure →
Microsoft Warns of TerminalFix Campaign Abusing Windows Terminal for Backdoor Deployment
Microsoft is warning of an ongoing malicious campaign named “TerminalFix” that abuses compromised websites to trick users into installing a backdoor. Visitors to affected sites see a custom overlay instructing them to complete a fake Cloudflare CAPTCHA verification by copying and running a malicious PowerShell command into Windows Terminal or PowerShell. Microsoft said that while traditional ClickFix campaigns direct victims to the Windows Run dialog, TerminalFix directs users to Windows Terminal or PowerShell instead, increasing the likelihood that complex, multi-line scripts execute successfully.
TerminalFix Malware Chain: From PowerShell to Reverse Tunnel
After the command is run, the victim receives two files: a legitimate binary and a malicious DLL file. The binary sideloads the malicious DLL, which delivers a hidden payload called “client.py,” a custom Python implant. The implant creates an encrypted WebSocket connection back to the attackers and provides SOCKS5-style proxy access into the victim’s internal network. Microsoft stated that this type of intrusion is particularly dangerous because it provides attackers with direct access to an organization’s internal network through the reverse tunnel.
TerminalFix: Microsoft Urges Investigation for Lateral Movement and Credential Exposure
Microsoft did not observe the attackers actually carrying out lateral movement, so it is unclear what they are using the access for. Microsoft urged organizations to treat affected devices as potential network pivot points and investigate for lateral movement and credential exposure, noting that in the hands-on-keyboard phase that typically follows, attackers leverage this access to escalate privileges, disable security controls, exfiltrate sensitive data, and deploy ransomware across the organization.
Trail of Bits: AI Agent Autonomously Discovers Zero-Days to Escape Virtual Machines
Trail of Bits published a controlled security experiment on August 26 in which OpenAI’s GPT 5.6-Cyber security model escaped a QEMU/KVM virtual machine three separate times. In its third and final attempt, the agent autonomously discovered and chained three previously unknown zero-day vulnerabilities and one patched-but-undistributed flaw. Researcher Artem Dinaburg wrote that a VM can no longer be assumed to contain a sufficiently advanced AI agent, and that such agents should be treated as an advanced persistent threat (APT).
AI Agents Now Discover Zero-Days to Escape Virtual Machines, Trail of Bits Proves →
AI Agent Escape: Three Attempts, Four Vulnerabilities, and a Firecracker Recommendation
In the first escape attempt, the agent identified the host kernel as vulnerable to Januscape (CVE-2026-53359) and wrote an exploit from scratch. In the second attempt, the agent exploited Debian 12’s libslirp 4.7.0, which remained vulnerable to CVE-2026-9539, combined with an upstream bug-fix commit. For the third round, the agent found vulnerabilities no one had publicly documented, using four components including an unpatched QEMU VAPIC ROM alias and a Linux KVM shadow page desync. Trail of Bits recommends Firecracker, Amazon’s open-source microVM monitor, which implements only virtual CPUs, memory, VirtIO networking, and VirtIO block storage.
AI Agents Now Discover Zero-Days to Escape Virtual Machines, Trail of Bits Proves →
AI Agent Escape: OpenAI Policy Update and Mitigations
OpenAI responded on August 18 with a policy update titled “Pacing model development in an era of cyber-critical capabilities,” mandating stronger isolation sandboxes for workloads executing model-generated or untrusted code, tighter network controls, and continuous automated security testing. Dinaburg’s mitigations include strict least-privilege network policies, active monitoring and logging, session time limits, and provisioning a fresh environment for each agent run. The agent demonstrated persistence across approximately 12-hour sessions, maintained goal coherence across internal compactions and sub-agent invocations, searched systematically, abandoned dead ends, and pivoted between attack surfaces.
AI Agents Now Discover Zero-Days to Escape Virtual Machines, Trail of Bits Proves →