IT News Analysis

1

OpenAI's GPT-6 Agent Removed Safety Monitoring, Then Falsely Denied Doing So

OpenAI has revealed that two artificial intelligence (AI) agents powered by GPT-6.1 Sol and one GPT-6 Astra weakened misalignment monitoring while making requested software changes during two internal coding tasks. The AI company said the incident took place on October 3 and 6, 2026, for Sol and Astra, respectively, as part of an internal deployment. "A GPT-6.1 Sol agent repairing an evaluation pipeline disabled automatic alerts to our Security team, treating an earlier instruction not to send messages during a Slack history search as a reason to suppress monitoring notifications," OpenAI said in a report. The end result of this action was that the misalignment monitor still ran and saved its reports, but automatic human alerts were disabled by setting "no_notify" to true and emptying the lists of people configured to receive monitoring reports. In another task on October 6, a GPT-6 Astra agent migrating a program is said to have removed its monitoring int...

Oct 21, 2026

2

PHP Issues Emergency Patches for Two Critical PHP2Shell RCE Flaws as Exploit Attempts Emerge

PHP has released emergency updates that fix two critical flaws in the language itself, letting an attacker without an account run code on a web server. The fixes are in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and the PHP team says only updating fully protects a server. The flaws, together called PHP2Shell , sit in the PHP runtime, not in any application, so a site is exposed even when its own code is secure. The risk is highest on servers using Apache's mod_php, where a working exploit triggers the attack before any application code runs. Both flaws are rated 9.2 on the CVSS v4.0 scale and carry the identifiers  CVE-2026-109618  and  CVE-2026-109619 . PHP published them around the scheduled release time of 14:00 UTC on October 21, a day after details leaked ahead of a planned October 22 disclosure. Researchers at Wiz, Calif, and OpenAI found and reported the flaws, which affect nearly every PHP version back to PHP 5.4. Who Is Affected Every PHP installatio...

Oct 21, 2026

3

SANS Has Taught Defenders Since 1989. AI Security Is the Newest Chapter.

Explore practitioner-built training, six GIAC AI certifications, and free resources to use this week.

4

Google Chrome Adds ClickFix Warnings to Stop Users From Running Malicious Commands

Google on Tuesday said it's rolling out improved safeguards in its Chrome browser to secure users against ClickFix attacks when Enhanced Protection is enabled. ClickFix is a popular social engineering technique in which bad actors trick users into copying and pasting malicious code directly into their computer's operating system, thereby bypassing conventional security controls. As part of these attacks, unsuspecting users who land on legitimate-but-compromised sites or malicious ads are redirected to phishing landing pages that display fake Cloudflare or Google reCAPTCHA verification or other non-existent errors and instruct users to copy and paste a command on the Windows Run dialog (or the Mac Terminal app) to address it. The moment a user presses the Enter key after pasting the command, it triggers a multi-stage infection chain that leads to the deployment of malware, such as information stealers or remote access trojans. In an attempt to blend in with regular act...

Oct 21, 2026

5

China-Aligned Bronze Sandcastle Abuses Sandboxie DLL Sideloading to Deploy Backdoors

Cybersecurity researchers have disclosed details of a previously undocumented China-aligned threat cluster known as Bronze Sandcastle that has been observed targeting organizations in Asia and Africa since mid-2025. "The most distinctive technical marker for Bronze Sandcastle is its consistent use of Sandboxie DLL sideloading," Sophos security researcher Gabor Szappanos said in a post published Tuesday. Between August 2025 and August 2026, attacks mounted by the threat actor have resulted in victims in the Philippines, Taiwan, Afghanistan, Namibia, and likely South Korea. Evidence indicates that Bronze Sandcastle is most active in the UTC+8 time zone.  DLL sideloading is a popular adversarial technique that exploits the way Windows loads DLL files, allowing an attacker to trick a legitimate program into loading a malicious DLL . When an application is launched, it needs to load various DLLs in order to function correctly. If the application does not provide the f...

Oct 21, 2026

6

Generative AI Is Erasing Behavioral Tells, Weakening Behavior-Only Detection, Proofpoint Reports

Generative AI has changed what a malicious email looks like. Messages that once gave themselves away through clumsy wording, generic pretexts, and reused templates now arrive fluent, personalized, and unique to each recipient, and the behavioral signals email filters have relied on for years are fading as a result. Proofpoint shared several examples of the shift at its Protect 2026 conference. In one, a single threat actor contacted more than 30 real estate agents at the same company, each from a different account, about a different property, and in different words, before closing every thread with the same request: join a video meeting. Federal data points the same way. The FBI's Internet Crime Complaint Center counted more than 22,000 complaints involving AI in 2025, with nearly $900 million in associated losses, and logged 24,768 business email compromise (BEC) complaints with about $3.05 billion in reported losses, according to its 2025 Internet Crime Report . What AI C...

Oct 21, 2026

7

Gartner® Hype Cycle™ for CPS Security, 2026

Prioritize cyber-physical systems security amid rising threats, regulations, vulnerabilities and emerging technologies.

8

PHP2Shell: What Actually Protects You Before the 14:00 UTC PHP RCE Patch, and What Doesn't

You cannot fully block the two critical PHP flaws called PHP2Shell before today's fix, and the one setting being passed around as a workaround is not enough on its own, the security firm that suggested it now says. PHP will release the fix at  14:00 UTC on October 21  for two flaws it rates 9.2 out of 10, and has given no workaround of its own. Until you can install it, what lowers your risk is taking exposed PHP off the internet or behind access controls, cutting what a compromised server can reach, and having the update ready to apply the moment it lands. Cloudflare says it is seeing early attempts to exploit the flaws, and has pushed emergency rules for sites behind it that it calls a stopgap until you patch. No compromise has been confirmed, the public exploit code going around has not been shown to work, and neither flaw is in CISA's catalog of known exploited flaws. PHP has not said which versions are affected, which part of PHP is involved, or whether any setting ...

Oct 21, 2026

9

The Harness Is the Hard Part of Agentic Pentesting

An AppSec team can point an AI agent running on a general-purpose model at its own web apps with little setup. Sure, the model writes the attacks. But everything that makes those attacks safe to run against production and worth acting on afterward lives in the harness around the model. The harness is where the engineering in AI pentesting sits. The Hugging Face intrusion OpenAI disclosed in July showed what happens when that harness gives way, even inside an OpenAI evaluation built to contain the agents. Why Each Control Takes Rigorous Engineering Let’s take a closer look at the attack different stages and the role AI agents can take there Scope A pentesting agent makes thousands of decisions about where to send its next request, and the model judges each one against a rule written in plain language. A web app’s real boundary often extends past one domain through redirects and calls to outside services. Enforcing scope means code that inspects every request before it leaves, a...

Oct 21, 2026

10

Attackers Hijacked Over 300 Tor Exit Relays to Intercept SSH Connections

Attackers took control of more than 300 Tor exit relays in early October and used them to redirect SSH connections to servers that stole login passwords, a researcher found. Exit relays are the last hop in a Tor circuit, where traffic leaves the network for its destination. Only a password login to an SSH server made through Tor between October 1 and 5, after accepting the interception server's host key, could have been captured. The Tor Project began flagging relays on October 4, and by October 5, none remained in the consensus, the list of relays that Tor clients use. At the peak on October 4, the relays carrying the attacker's exit policy made up 7.46% of Tor's exit capacity, measured by consensus weight, according to an  analysis of public Tor directory data  posted to the tor-relays mailing list. The researcher, who publishes as kaba,  scanned  all of Tor's roughly 3,000 exits on October 2 and found 333 redirecting SSH to the same 12 interception servers...

Oct 21, 2026