Tech News, October 15: FBI Arrests an OpenAI Engineer, Cisco’s 9.8 Zero-Day, and Netflix’s 1,000 Layoffs

The last 24 hours have been unusually dense. The FBI arrested an OpenAI engineer accused of planting a covert listening device inside the company’s San Francisco offices. Cisco disclosed an actively exploited zero-day rated 9.8 in its email gateway. Researchers published details of a zero-click credential-theft flaw in Claude Code Desktop. Netflix confirmed roughly 1,000 layoffs, and Anthropic moved closer to a November IPO after telling investors it made $16.8 billion in quarterly revenue.

FBI arrests an OpenAI engineer over a hidden recording device

The Justice Department said Tuesday that federal agents arrested Abdulrahman Alfozan, 35, a Saudi national who had worked as an OpenAI engineer since July 2025. He is charged with installing a hidden audio recording device in an executive conference room at one of OpenAI’s San Francisco buildings.

OpenAI found the device on September 30, plugged into a USB port on the back of a television. According to the FBI complaint, the thumb-sized device had a microphone, could connect to a camera, was concealed in toilet paper, and was installed on or about September 28. Its stored data was encrypted and inaccessible to investigators. A second, similar customized device was later found in Alfozan’s personal bag.

OpenAI said it used its own internal AI models to identify the person who planted the device and trace where it came from, and that it has evidence he intended to place additional devices in sensitive areas to secretly record meetings. The company terminated him and said it has “no indications” of model weights, source code, or customer data being exfiltrated. The FBI investigation is ongoing.

Why it matters: frontier AI labs are increasingly treated as national-security assets, and this case shows the threat model includes the building itself, not just the network. It is also a strange inversion: an AI company used its own models as a forensics tool against an insider. Whether the encrypted data yields anything further is now the FBI’s problem to solve.

Cisco patches an actively exploited 9.8 zero-day in its email gateway

Cisco published a Wednesday advisory for CVE-2026-76519, a critical flaw in the Secure Email Gateway’s Advanced Malware Protection feature. A relative path traversal weakness in how the gateway handles email attachments lets an unauthenticated attacker send a crafted email through the appliance, write arbitrary files, and execute commands with root privileges. Cisco scores it CVSS 9.8 and says there are no workarounds. Its product security incident response team became aware of active exploitation in October 2026.

A gateway is only exposed when two conditions hold: the File Reputation service is enabled and assigned to a mail policy, and the Content Scanner Tools package is older than version 25.3.0.5525. Admins can check exposure with the contentscannerstatus CLI command and the “Enable File Reputation” checkbox under Mail Policies in the management interface. Cisco has already applied mitigations for its cloud-hosted Secure Email Cloud customers and contacted appliance owners where it found evidence of compromise. Shadowserver tracking puts roughly 300 gateway instances exposed on the internet.

Why it matters: this is the second exploited Secure Email Gateway zero-day in about a month, after September’s CVE-2026-76461. Email gateways sit at the network edge with broad trust, and root access on one lets an attacker erase local logs and move laterally. If a virtual appliance shows signs of compromise, Cisco advises rebuilding the instance rather than upgrading, because an upgrade destroys forensic evidence.

“SneakPeek”: a zero-click flaw that could steal developer credentials from Claude Code

Noma Labs disclosed on October 14 a vulnerability chain it calls SneakPeek in Anthropic’s Claude Code Desktop application, rated CVSS 8.7. The flaw lived in the app’s automated HTML preview pane, an Electron-based Chromium panel that renders real-time previews of HTML files Claude generates or edits.

The attack worked like this: an adversary publishes a repository containing an HTML file larger than 256KB with an embedded script. When a developer asks Claude Code to review or edit the file, the preview pane renders it automatically. Noma found that files under 256KB were converted to data URLs with a strict Content Security Policy, while larger files were written to disk and loaded as file:// URLs, bypassing the CSP, gaining direct local filesystem access in the renderer, and running without request filtering. The script could then silently read SSH private keys, AWS and GitHub credentials, API keys, Kubernetes configs, and Claude’s own OAuth tokens, with no click or prompt from the user.

The disclosure timeline is long: Noma reported it on July 4, Anthropic deployed an initial mitigation, researchers found a Web Worker bypass on July 24, and the final fix shipped August 17 in Claude Desktop 1.30096.1 and later. The published reporting describes a resolved flaw, not evidence that any particular user was compromised.

Why it matters: AI coding tools routinely render untrusted content, and the sandbox boundary turned out to be a preview pane. Developer machines hold the keys to production, so a zero-click path from a poisoned repo to credential exfiltration is exactly the kind of flaw that keeps security teams up at night. If you use Claude Code Desktop, confirm you are on 1.30096.1 or later.

Netflix cuts about 1,000 jobs, its biggest round since 2022

The Wall Street Journal reported Wednesday that Netflix is cutting roughly 1,000 jobs, about 6% of its workforce, in an email to staff from co-CEOs Ted Sarandos and Greg Peters. It is the streamer’s largest round of layoffs since 2022, when it cut around 450 jobs. The reductions hit product, technology, legal, and human-resources teams hardest. Netflix had about 16,000 full-time employees at the end of 2025.

The context is consolidation and slowing momentum. Paramount completed its $110 billion acquisition of Warner Bros earlier this month, forming the Skydance media giant Netflix lost the bidding war for. The company’s stock has traded well below its 2026 peak as investors question user engagement growth, and management framed the cuts as discipline about “where we prioritize our people and resources.” Netflix reports third-quarter earnings on October 20.

Why it matters: the memo targets the teams that build and support the platform, which creates an execution risk even as it protects margins. The earnings call next week will show whether this is cost discipline or the start of a harder stretch for streaming’s biggest player.

Anthropic holds its investor day as a November IPO comes into view

Anthropic met prospective institutional investors at its San Francisco headquarters on October 14, advancing plans for what could be the largest IPO in history. Bloomberg reports the company could begin formal marketing as soon as the week of November 9 and start trading before Thanksgiving, with investors discussing a $1.8 trillion to $2 trillion valuation. OpenAI, by contrast, has pushed its own listing to 2027 while raising at least $30 billion at a roughly $1.4 trillion valuation.

Separately, the Financial Times reported that Anthropic told investors its third-quarter revenue hit $16.8 billion, up 45% sequentially from $11.6 billion in Q2 (and $4.8 billion in Q1), and that it turned profitable on an adjusted basis for the first time. The company raised $65 billion in May at a valuation near $965 billion. Anthropic also opened a Singapore office this week and signed an AI safety MOU with the Singapore government.

Why it matters: the revenue numbers, if they hold up in a public filing, explain the IPO timing. Anthropic’s pitch is enterprise adoption of Claude, and a November listing would land it on public markets before OpenAI. Until an S-1 appears on EDGAR, the $2 trillion figure is a discussion target, not a fact.

The week’s flaws at a glance

Key takeaways

• Patch your email gateway: if Cisco Secure Email Gateway has File Reputation enabled, update Content Scanner Tools to 25.3.0.5525 or later, and check for signs of compromise first. Rebuild rather than upgrade if compromise is suspected.
• Update Claude Desktop: anyone using Claude Code Desktop should be on 1.30096.1 or later, the version with the final SneakPeek fix.
• Watch October 20: Netflix’s Q3 earnings will show whether the 1,000 layoffs are margin discipline or a response to deeper engagement problems.
• Watch early November: Anthropic’s IPO marketing could start the week of November 9; the public S-1 filing is the moment the $2 trillion valuation gets real or doesn’t.
• Insiders are the story this week: a planted device at OpenAI, an AI-assisted internal investigation, and two zero-days found by the vendors themselves before attackers moved faster anyway.

References

• Reuters (via SAMAA TV): FBI arrests Saudi OpenAI engineer over covert recording device
• BleepingComputer: Cisco patches actively exploited Secure Email Gateway zero-day
• NeoTeo: Noma says Claude Desktop 1.30096.1 fixed Claude Code’s SneakPeek flaw
• LDS: Anthropic patches Claude Code preview vulnerability
• The Hindu BusinessLine (Reuters): Netflix cuts about 6% of jobs in biggest layoff since 2022, WSJ reports
• The Hindu BusinessLine (Bloomberg): Anthropic is said to plan pre-IPO investor day as listing nears
• OOSSA (Financial Times): Anthropic reports $16.8 billion Q3 revenue, up 45% sequentially