Report · drew.silcock.dev

Is this the new cyber normal?

drew.silcock.dev

Source published October 23, 2026

Summary: 2 min read

🔥🔥🔥🔥🔥

58/100

Summary

A cluster of critical security disclosures in October 2026 highlighted how quickly vulnerabilities can move from discovery to attempted exploitation, particularly when AI models help researchers and attackers analyze code and develop proofs of concept. PHP2Shell, comprising CVE-2026-109618 and CVE-2026-109619, involved remote-code-execution flaws in PHP reportedly found by Calif and Wiz using GPT-5.6 Sol and Gemini 4 Argon. A leak accelerated public awareness before patches were available, and AWS reported observing attempts to exploit the flaws from October 17. Wiz estimated that affected PHP deployments could cover more than 70% of the internet. Other incidents included Mozilla’s October 20 blocking of eight eMudhra intermediate certificates after weak private keys were reported; eMudhra later said it would exit the global SSL/TLS certificate business. Railway disclosed possible access to and decryption of customer environment variables, while GitLab issued critical patches for multiple CVSS 10.0 code-execution flaws in October. LiteLLM said its compromised version 1.107.0 PyPI release and Docker images contained credential-stealing malware after an attacker used GitHub CLI credentials to run CI jobs. PingCAP disclosed unauthorized access to cloud resources as FulcrumSec claimed to have exfiltrated 10.6 TB of data. The incidents do not all have confirmed AI involvement, but they underscore shrinking response windows and the importance of patching, secret management, and coordinated vulnerability disclosure.

Key Takeaways

PHP2Shell refers to CVE-2026-109618 and CVE-2026-109619, two PHP remote-code-execution vulnerabilities reportedly discovered with help from frontier AI models. AWS said it observed attempted exploitation from October 17, before public disclosure.
Mozilla blocked eight eMudhra intermediate certificates on October 20, 2026, after a researcher reported severely weak private keys. eMudhra subsequently announced plans to leave the global SSL/TLS certificate business.
GitLab released patches for multiple maximum-severity code-execution vulnerabilities in October 2026, including flaws in CI/CD regular-expression parsing. The reported CVSS 10.0 issues were submitted through GitLab’s HackerOne bug bounty program.
LiteLLM said version 1.107.0 on PyPI and associated Docker images were compromised with credential-stealing malware. The company attributed the incident to an attacker’s access to GitHub CLI credentials used to run CI jobs.
PingCAP reported unauthorized access to cloud resources, while the extortion group FulcrumSec claimed it exfiltrated 10.6 TB of data; PingCAP said it found no evidence that data stored in TiDB was affected.

What the discussion said

Commenters largely treated the incident as evidence that AI is compressing the security timeline: models can search enormous codebases, assemble proof-of-concept attacks, and turn a disclosed flaw into a practical threat faster than conventional patch cycles can respond. The immediate fear is not merely more bugs, but the collapse of the buffer defenders once had between discovery, disclosure, and mass exploitation. Several readers argue that attackers can accept reckless automation while defenders must validate fixes, coordinate releases, and protect production systems. There was genuine respect for the defensive upside. People described LLMs uncovering overlooked flaws in mature personal code, helping audit legacy systems, and even modernizing vulnerable code cheaply and quickly. Some see a plausible end state where automated continuous scanning, broad dependency maintenance, and more memory-safe or formally verified software drain the backlog of old defects. But the dominant skepticism is that AI will not simply find a fixed stock of human mistakes and then make software safe. Fast feature churn creates fresh attack surface, generated code can omit basic controls or repeat subtle weaknesses, and an auditor model finding flaws in another model's output proves neither is reliable by default. A few readers also challenged whether the reported exploitation wave was as broad as advertised, but that did little to soften concern about the underlying trend.

Where opinion split

The central dispute is whether AI-driven vulnerability discovery is a temporary cleanup phase or a durable escalation of cyber risk. Optimists argue that a finite backlog of human-written defects can be exhausted and then continuously scanned and repaired by capable models. Skeptics counter that software keeps changing, AI-generated code adds new flaws, and automation favors attackers because safe remediation remains slower, costlier, and more cautious than exploitation.

Read original article