A little too OpenAI.

An OpenAI engineer faces charges over an alleged covert recording device. Attackers exploit a PAN-OS firewall flaw and an Ivanti zero-day. Aoki infostealer hunts AI credentials. Researchers uncover critical weaknesses in an AI gateway. CikaBika Bot turns Discord into a command center. Online ads deliver mercenary spyware without a click. Microsoft disrupts a massive Chinese-language phishing operation. And INTERPOL leads an international crackdown on child sexual abuse material. Character.ai runs out of tokens.

Today is Thursday October 15th 2026. I’m Dave Bittner. And this is your CyberWire Intel Briefing.

An OpenAI engineer is charged with allegedly planting a covert recording device. 

A 35-year-old OpenAI engineer has been arrested and charged with allegedly planting a covert recording device in an executive conference room at the company’s San Francisco offices. Federal prosecutors say Abdulrahman Alfozan, a Saudi national employed by OpenAI since July 2025, concealed an encrypted recording device inside a television’s USB port, wrapped in toilet paper. Investigators say surveillance footage placed Alfozan near the room, while chatbot queries associated with his spouse’s account included instructions for building and concealing audio recorders. Security personnel subsequently discovered a second recording device in his unattended bag. The conference room reportedly hosted sensitive discussions involving proprietary research, product development, and government relationships. Authorities have not disclosed whether any conversations were successfully intercepted. Alfozan faces up to five years in prison if convicted.

A high-severity PAN-OS firewall vulnerability is under active exploitation. 

Attackers are actively exploiting a high-severity vulnerability in Palo Alto Networks’ PAN-OS firewall software, prompting an urgent federal response. The flaw, tracked as CVE-2026-0322, allows authenticated attackers with network access to execute code with root privileges. Palo Alto Networks says exploitation has been very limited, but CISA has ordered federal agencies to patch affected systems and investigate potential compromises by October 17. The vulnerability affects PA-Series and VM-Series firewalls, along with Panorama management systems. Patches are available for supported versions, but no complete workarounds exist. Researchers warn that patching alone isn’t enough. Organizations should review administrative logins, investigate suspicious configuration changes, and restrict management interfaces to trusted networks. Once attackers control a firewall, they potentially control the very device responsible for protecting the network.

CISA urges immediate patching of an Ivanti Connect Secure zero-day.

CISA is urging organizations to patch Ivanti Connect Secure appliances following active exploitation of a critical zero-day vulnerability. The flaw, tracked as CVE-2026-104935, allows attackers to bypass authentication and execute commands remotely. Researchers at Volexity say an unidentified threat actor began exploiting the vulnerability October 3, chaining it with additional weaknesses to gain root access. Attackers harvested domain credentials, searched for administrator passwords and BitLocker recovery keys, and probed Active Directory. They also covered their tracks by manipulating memory and deleting logs, potentially evading Ivanti’s Integrity Checker Tool. Ivanti released patches October 13, though a separate privilege escalation flaw remains unpatched. CISA warns that patching alone isn’t sufficient. Organizations should investigate potential compromises, rebuild affected appliances, and rotate exposed credentials and certificates.

Aoki infostealer targets AI credentials. 

A relatively new infostealer called Aoki is expanding its appetite for stolen credentials, and AI coding assistants are now on the menu. According to Symantec, newer versions of the malware target authentication tokens, session files, and API keys associated with tools including Claude, ChatGPT, Codex, GitHub Copilot, Cursor, and Gemini. Earlier versions focused primarily on browsers, cryptocurrency wallets, and password managers. Researchers warn that stolen AI credentials could allow attackers to impersonate developers, run up API charges, and potentially access connected code repositories, cloud services, and development pipelines. Symantec says the malware, first observed earlier this year, remains under active development. The researchers haven’t identified the group responsible, but Aoki’s continued focus on cryptocurrency suggests its operators are financially motivated.

Researchers uncover high-risk vulnerabilities in a popular open-source gateway. 

Researchers at Check Point have uncovered two high-risk vulnerabilities in LiteLLM, a popular open-source gateway that connects enterprise AI applications to providers including OpenAI, Google, and Anthropic. The flaws, tracked as CVE-2026-35029 and CVE-2026-59823, could allow ordinary users to gain administrative control of the gateway. One vulnerability could enable remote code execution and potentially complete system takeover. The other could expose API keys through manipulated outbound requests. That’s particularly concerning because AI gateways often centrally manage credentials, permissions, and access to multiple AI services. Check Point says any authenticated user could potentially exploit a vulnerable deployment. The flaws affect LiteLLM versions earlier than 1.83.9, which includes fixes for both issues. Organizations are urged to update promptly.

CikaBika Bot uses Discord for command and control. 

Researchers at Darktrace have uncovered a malware campaign distributing CikaBika Bot, a Rust-based remote access trojan that uses Discord for command and control. The attack begins with a document-themed lure that launches hidden PowerShell commands and attempts to bypass Windows security protections. Once installed, the malware collects system information, captures screenshots, and allows attackers to execute commands, manipulate files, and conduct surveillance. It also uses Ngrok to exfiltrate stolen data. Darktrace says the campaign illustrates how legitimate online services can support malicious operations.

A surveillance operation uses ordinary online advertisements to deliver mercenary spyware. 

Researchers at the University of Toronto’s Citizen Lab have uncovered a surveillance operation that uses ordinary online advertisements to deliver mercenary spyware without requiring victims to click anything. Dubbed AdFection, the operation allegedly involves Israeli advertising company Discover-Tech, which researchers say facilitates surveillance for spyware vendors including Candiru. Citizen Lab linked the campaign to zero-click attacks targeting at least 45 devices at Al Jazeera. Discover-Tech’s ads appeared on more than 120,000 websites and 70,000 mobile apps between March and June, with activity also observed on sensitive government and international networks. Researchers also identified a related operation, dubbed DARK CARTOGRAPHER, that allegedly exploits mobile carrier vulnerabilities to connect advertising identifiers with phone numbers. Google and IAB Europe have removed Discover-Tech from their approved vendor lists. Citizen Lab warns that the findings expose fundamental security weaknesses in digital advertising, where the machinery built to target consumers can also be used to target them for surveillance.

 

Interpol leads an international effort against CSAM. 

INTERPOL says a coordinated law enforcement operation spanning 17 countries in Latin America and Europe has resulted in 95 arrests targeting the distribution of child sexual abuse material through online messaging groups. Dubbed International Effort Confluencia, the operation involved four investigations between March and September, with authorities executing 127 search warrants. Investigators rescued a five-year-old child whose parent had allegedly been producing and selling abuse material. In Panama, police arrested the administrator of a messaging group with more than 700 members. An Argentine suspect allegedly used sextortion against girls around the world. INTERPOL warns that messaging platforms increasingly facilitate grooming, exploitation, and the commercialization of abuse material. Authorities expect evidence seized during the operation to help identify additional offenders and children at risk.

Character.ai runs out of tokens. 

Character.ai’s bankruptcy offers a cautionary tale for the AI industry: attracting millions of users is one thing. Making money from them is quite another. Valued at a billion dollars in 2023, the AI companion company was spending as much as $13.8 million a month on computing costs, while subscription revenue never exceeded $2.8 million. Lawsuits involving alleged harm to users added substantial legal expenses, but the company acknowledges its business was unsustainable even without them. Character.ai tried cutting costs by simplifying its models, laying off employees, introducing advertising, and charging for additional features. Users got a diminished experience, and the company still lost eight to nine million dollars a month. The service remains operational during bankruptcy proceedings. Meanwhile, larger competitors can draw on deep-pocketed investors or established businesses to keep funding their AI ambitions. Character.ai discovered that impressive technology and enthusiastic users don’t necessarily add up to a viable business.

And that’s the CyberWire.

For links to all of today’s stories, check out our Daily Briefing at the cyberwire dot com.

We’d love to know what you think of this podcast. Your feedback ensures we deliver the insights that keep you a step ahead in the rapidly changing world of cybersecurity. If you like the show, please share a rating and review in your podcast app. Please also fill out the survey in the show notes or send an email to cyberwire@n2k.com

N2K’s lead producer is Liz Stokes. We’re mixed by  Tré Hester, with original music by and sound design Elliott Peltzman. Our contributing host is Maria Varmazis. Our executive producer is Jennifer Eiben. Peter Kilpe is our publisher. And I’m Dave Bittner. Thanks for listening.