Latest cybersecurity reporting from selected sources.

Google Says Criminals Used AI to Discover and Code ExploitA cybercriminal group came close to launching a mass attack earlier this year, armed with a software exploit that an AI model had built from scratch, said Google researchers. Google said it worked with the affected vendor to patch the flaw before an attack could be launched.

Zero-Click Vulnerability Let Attackers Weaponize Enterprise AI AssistantGoogle patched a vulnerability in Gemini Enterprise that allowed attackers to steal corporate data through a shared document, calendar invitation or email without any user action or security alerts. No malware was executed, no credentials were phished and no data left through approved channels.

Also, North Korean Hackers Remotely Wipe Android DevicesThis week, the U.K. government probed Chinese electric buses for a kill switch, APT37 abused Google's Find Hub in South Korea, Conduent said its January hack will cost it more, Hyundai disclosed a breach and Patch Tuesday. OWASP added two new categories to its Top 10 web application vulnerabilities.

Security Experts Laud Project Zero's Push for Greater Transparency, Faster PatchesGoogle is trying out a new approach to publicizing flaws found by its in-house bug hunters meant to get patches more rapidly into end users' hands. Under a trial policy effective immediately, Google's Project Zero team will publish a general alert to the public within seven days.

Hacking Group UNC6148 Steals Credentials With New OVERSTEP Rootkit, Google SaysA cybercrime group used a backdoor in a fully patched SonicWall appliance to steal credentials and may have sold the stolen data to ransomware groups as part of an ongoing campaign, Google Threat Intelligence Group found. The firm attributed the campaign to a cybercrime group it tracks as UNC6148.

Also, Oracle Denies Cloud Breach, Blames Hack on Obsolete ServersThis week, Port of Seattle notified victims, Oracle blamed hack on obsolete servers, Google and Microsoft released April patches, WK Kellogg breached, six arrested in Spain for AI-investment scam, Scattered Spider's "King Bob" pleaded guilty, SmokeLoader users busted.

Also: Gootloader Malware, GCHQ Intern Pleads Guilty, Check Point Breach UpdateThis week, a "Fast Flux" warning, Gootloader malware, an GCHQ intern pleaded guilty to stealing top secret data and Check Point undercuts hacking claim. Also, Google rolled out end-to-end encryption for some Gmail users, Apple backported patches and Dutch prosecutors cut internet access.

Also: Google Fixes YouTube Vulnerabilities That Could Have Exposed User EmailsThis week: Microsoft, Ivanti and Google release fixes for critical vulnerabilities and urge priority patching; Lee Enterprises confirms a cyberattack disrupted newspaper operations; and thousands of KerioControl Firewalls exposed to critical remote code execution flaws.

SysTrack LsiAgent Installer Flaw Escalates Privileges LocallyA flawed Microsoft software installer application developed by Lakeside Software could enable attackers with lower privileges to gain full system access. The local privilege escalation vulnerability uncovered by Google Mandiant has since been patched.

Now-Fixed Flaw Is Big Sleep's First Real-World Bug Find, Say ResearchersGoogle's "highly experimental" artificial intelligence agent Big Sleep has autonomously discovered an exploitable memory flaw in popular open-source database engine SQLite. The researchers detail how the AI agent discovered the now-patched vulnerability.

Attackers Could Exploit Flaw to Run Malicious Code on Google' s, Customers' ServersGoogle patched a critical remote execution vulnerability in its cloud platform Cloud Composer service, "CloudImposer," which could have allowed attackers to compromise millions of servers, say researchers from Tenable. The CloudImposer vulnerability could lead to the Jenga Tower effect.

Also: Turla Targets European Missions and Google Patches Chrome Zero-DaysThis week, hackers used a Linus backdoor and a Microsoft client management tool; Santander Bank, the Helsinki Education Division, an Australian energy provider and auction house Christie's were breached; hackers targeted European missions in the Middle East; and Google patched a zero-day flaw.

Also: Catching Up With Spain's Most Dangerous HackerThis week, the FCC OK'd cybersecurity labeling, DarkGate exploited Google, Fortinet patched a bug, cyberattacks hit the French government and employment agencies, Google restricted Gemini AI chatbot and paid bug bounties, Microsoft had Patch Tuesday, Marine Max was attacked, and Alcasec moved on.

Google Fixes Out-of-Bounds Memory Access Flaw, Microsoft Edge Browser Also AffectedGoogle released an urgent fix for the first zero-day vulnerability of the year in its Chrome web browser, warning the bug is under active exploitation. Google blamed an out-of-bounds memory access flaw in its V8 JavaScript rendering engine. It also affects Microsoft Edge browser.

Zimbra Patched the Cross-Site Scripting Vulnerability on July 25A zero-day flaw in the Zimbra Collaboration email server proved to be a bonanza for hackers as four distinct threat actors exploited the bug to steal email data and user credentials, says Google. Most of the exploit activity occurred after Zimbra had posted a hotfix on July 5.