🎧 Nov 19 Brief · archive
Welcome to the BrevFeed daily tech briefing for Thursday, November 19. Here are today's top 3 stories in tech. Let's get into it.
In security, the PHP team has issued an urgent warning about two critical remote code execution vulnerabilities, identified as CVE-2026-109618 and CVE-2026-109619. These flaws, rated 9.2 on the CVSS v4.0 scale, affect PHP versions 5 through 8, which power about 70% of all websites.
Emergency security updates were released on October 21st, ahead of schedule, due to information about the vulnerabilities reportedly leaking. The U.S. National Security Agency and CISA have both issued warnings, with CISA adding the vulnerabilities to its Known Exploited Vulnerabilities catalog and ordering federal agencies to patch by Thursday.
The vulnerabilities, which have been present in PHP since version 5 in 2004, were discovered in mid-August by researchers from Wiz, Calif, and OpenAI, with Google's Gemini 4 Argon AI model also contributing. Exploitation attempts have been observed in the wild, with some threat actors using the flaws to compromise firewalls and deploy malware.
While many alleged exploits generated by large language models have not functioned against vanilla PHP instances, successful compromises have been recorded. Organizations like Cloudflare and Automattic have deployed protections, and the UK's NCSC urged organizations to apply fixes or take systems offline, with 32 UK government services being affected.
In security, a new social engineering tactic called "ClickFix" is rapidly gaining traction among cybercriminals. This method tricks users into manually executing malicious commands by presenting fake CAPTCHAs or error messages, effectively bypassing traditional security measures.
ClickFix attacks have seen a 517% surge from late 2024 into early 2025, targeting a wide range of systems including Microsoft 365 accounts and Mac users. Attackers leverage compromised websites to inject JavaScript that delivers these prompts, often leading to the installation of various malware like cryptominers or information stealers.
The technique is particularly concerning because it exploits user habits rather than system vulnerabilities, making it difficult for conventional defenses to detect. Microsoft has noted ClickFix as the most common initial access method last year, accounting for 47% of their incident notifications. This highlights the urgent need for enhanced user education and adaptive security strategies to counter these evolving threats.
In AI, Google has unveiled Gemini 4 Argon, its new flagship AI model, designed for deep reasoning across complex workflows. This model boasts a one million output token limit, a significant increase from previous versions, allowing for much longer and more intricate tasks.
Gemini 4 Argon shows advanced capabilities in coding, knowledge work, cybersecurity defense, and creative writing. Google reports it surpasses competitors like Claude Opus 5.5 and GPT-6 Astra in several benchmarks, including real-world software engineering and cybersecurity. It's also noted for its ability to operate without cyber guardrails for trusted defenders.
The rollout of Gemini 4 Argon is phased, starting with trusted cyber defenders and internal Google teams through the Fairwind Program. Access will then expand to paid API customers, AI Ultra subscribers, developers, and enterprises. Internally, thousands of Googlers are already using it for tasks like codebase migrations and data center memory optimization.
That's the BrevFeed daily briefing. We'll be back tomorrow with the stories that matter in tech. Thanks for listening.