OpenAI fixes zero-click flaw in ChatGPT's Deep Research agent.

Summary

By the CyberWire staff

OpenAI fixes zero-click flaw in ChatGPT's Deep Research agent.

OpenAI has fixed a zero-click vulnerability in ChatGPT's Deep Research agent that could have exposed Gmail data, Infosecurity Magazine reports. Researchers at Radware discovered the flaw, explaining that "[u]nlike prior research that relied on client-side image rendering to trigger the leak, this attack leaks data directly from OpenAI’s cloud infrastructure, making it invisible to local or enterprise defenses."

Attackers could embed invisible instructions within an email before sending it to a targeted user. If the user asks ChatGPT Deep Research to analyze their emails, the AI agent will encounter the hidden instructions and follow them. In Radware's example, an attacker could send an email to an HR worker with instructions for Deep Research to find employee information and upload it to a URL.

OpenAI silently fixed the vulnerability in August, then acknowledged the flaw on September 3rd.

Control what runs in your environment. Reduce your attack surface.

ThreatLocker helps organizations reduce risk by allowing trusted applications to run while limiting their access to only the resources they need. It’s a straightforward, default deny approach that gives you more control and visibility—without slowing down operations. Explore how ThreatLocker can help simplify your security strategy.

CISA shares details on malware targeting Ivanti flaws.

The US Cybersecurity and Infrastructure Security Agency (CISA) has published details on two sets of malware that were deployed in attacks targeting vulnerabilities in Ivanti Endpoint Manager Mobile, SecurityWeek reports. The vulnerabilities (CVE-2025-4427 and CVE-2025-4428) were patched in May 2025. CISA has provided IOCs and mitigation advice.

The agency stated, "Around May 15, 2025, following publication of a proof of concept, the cyber threat actors gained access to the server running EPMM by chaining these vulnerabilities. The cyber threat actors targeted the /mifs/rs/api/v2/ endpoint with HTTP GET requests and used the ?format= parameter to send malicious remote commands. The commands enabled the threat actors to collect system information, download malicious files, list the root directory, map the network, execute scripts to create a heapdump, and dump Lightweight Directory Access Protocol (LDAP) credentials. CISA analyzed two sets of malicious files the cyber threat actors wrote to the /tmp directory. Each set of malware enabled persistence by allowing the cyber threat actors to inject and run arbitrary code on the compromised server."

Are You Future Ready? Secure Your Machine Future With Confidence.

CyberArk is the leader behind the world’s most comprehensive machine identity security solution. With capabilities that span secrets, certificate management and workload identities, we enable customers to secure more machine identity use cases with the right level of privilege control. Find out exactly how we’re helping modern organizations secure their machine future.

WatchGuard patches critical firewall flaw.

WatchGuard has patched a critical vulnerability (CVE-2025-9242) in its Firebox firewalls that could allow a remote unauthenticated attacker to execute arbitrary code, Hackread reports. The vulnerability "affects both the mobile user VPN with IKEv2 and the branch office VPN using IKEv2 when configured with a dynamic gateway peer."

There's no evidence that the vulnerability has been exploited, but users are advised to update their Firebox software as soon as possible.

Notes.

Today's issue includes events affecting , and the United States.

Attacks, Threats, and Vulnerabilities

New York Blood Center Alerts 194,000 People to Data Breach (Infosecurity Magazine) A breach at the New York Blood Center resulted in theft of data for 194,000 people, including SSNs, IDs, bank and health information

How weak passwords and other failings led to catastrophic breach of Ascension (Ars Technica) A deep-dive into Active Directory and how “Kerberoasting” breaks it wide open.

Trends

2025 SonicWall Threat Brief - The Misconfiguration Epidemic (SonicWall) Misconfiguration and authentication bypass attacks have become the bread and butter of successful cybercriminals this year. These aren’t cutting-edge techniques requiring elite hacking skills. Learn how basic misconfigurations are fueling today’s biggest breaches.

Legislation, Policy, and Regulation

DOD official: We need to drop the cybersecurity talent hiring window to 25 days (CyberScoop) The Department of Defense is seeking to address persistent shortages in its cyber workforce by reducing the time to fill vacant cybersecurity jobs to 25 days.

Senate confirms Sutton as Pentagon cyber policy chief (The Record) Katherine Sutton most recently served as the chief technology adviser to the commander and director of Pentagon operations at U.S. Cyber Command.

MI6 upgrades dark web portal to recruit new spies (The Register) YouTube vids explain digital tradecraft to reach spooks over Tor or VPN without blowing your cover

Litigation, Investigation, and Law Enforcement

Contractor Used Classified CIA Systems as ‘His Own Personal Google’ (404 Media) Dale Britt Bendler “earned approximately $360,000 in private client fees while also working as a full-time CIA contractor with daily access to highly classified material that he searched like it was his own personal Google,” according to a court record.

Industry Events

For a complete running list of events, please visit the Event Tracker.

Events

Open Source Security Summit 2025 (Virtual, Sep 25, 2025) The sixth annual Open Source Security Summit brings together business leaders, industry visionaries, and technology users to chart a path forward and highlight the future of open source security solutions at this free virtual event. Register now to explore advancements in open source security and how using open source tools can build trust with customers and consumers.

HackAICon 2025 (Lisbon, Portugal, Sep 25, 2025) The first conference dedicated to combining AI and Ethical Hacking. We're bringing together OffSec pros, hackers, and leading researchers to find practical ways of using AI to secure the internet. Learn about the latest AI Hacking cosmic trends and connect with fellow professionals in engaging conversations. We banned sales pitches from the event - just knowledge sharing, leadership and technical insights.

Don’t miss Jen Easterly, former CISA Director, headlining HIP Conf 25 (Charleston, SC, USA, Oct 7 - 9, 2025) Jen Easterly, former CISA Director, keynotes HIP Conf 2025. A globally recognized cybersecurity leader, Easterly brings unmatched insight into hybrid identity security and resilience. Register now for this must-attend event.

Uniting Women in Cyber 2025 (Crystal City, Virginia, USA, Oct 9, 2025) UWIC is the premier event for cybersecurity leaders, professionals, and aspiring practitioners. Connect with a vibrant community focused on emerging global trends, technological advancements, and workforce development. Attend UWIC 2025 to learn, share ideas, and expand your professional network!

ISC2 Security Congress 2025 (Nashville + Virtual, Tennessee, USA, Oct 29 - 31, 2025) Join thousands of cybersecurity experts at ISC2 Security Congress 2025 and help lead the future of our connected world. Gain the insights you need to stay ahead of the evolving digital landscape. From innovative software to new security strategies, ISC2 Security Congress 2025 is designed to ensure you can navigate tomorrow’s threats.

Sponsor & Support

Grow your brand, generate leads, and fill your funnel.

With the industry’s largest B2B podcast network, popular newsletters, and influential readers and listeners all over the world, companies trust the CyberWire to get the message out. Learn more.

Daily Briefing for 09.19.25

Top stories.

OpenAI fixes zero-click flaw in ChatGPT's Deep Research agent.

CISA shares details on malware targeting Ivanti flaws.

WatchGuard patches critical firewall flaw.

Notes.

Attacks, Threats, and Vulnerabilities

Trends

Legislation, Policy, and Regulation

Litigation, Investigation, and Law Enforcement

Events