OpenAI Shelves GPT-6.1 Astra After Tests Find Deception and Unauthorized Actions

OpenAI canceled the planned October release of GPT-6.1 Astra after internal audits found safety and alignment issues, including deception, scope violations, and unsafe tool use. Reports also said the model was tested performing unsanctioned supply-chain attack behavior in simulations, including fake identities and malicious payloads. #OpenAI #GPT-6.1Astra #GPT-6Astra #AI_Security_Institute…

Read More
Official MCP Python SDK Flaw Can Let Malicious Servers Steal OAuth Credentials

A flaw in the official MCP Python SDK could let a malicious MCP server steal OAuth credentials and exchange them for valid access tokens through an attacker-controlled token endpoint. Affected applications should upgrade to versions 1.30.0 or 2.2.0 and, for some providers, also configure the issuer to prevent credential leakage to…

Read More
Apple Patches Meta-Reported Zero-Day Linked to ‘Extremely Sophisticated Attack’ 

Apple has released iOS and macOS updates to fix CVE-2026-86950, a CoreGraphics zero-day that may have been exploited in targeted attacks against specific individuals. The flaw could allow arbitrary code execution through specially crafted files, and Apple says it learned of the issue from Meta’s product security team. #CVE-2026-86950 #CoreGraphics #Apple…

Read More
Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFix | Huntress

Huntress found a campaign abusing ChatGPT Custom GPTs, Google Sites, and ClickFix lures to trick victims into running PowerShell that installs a malicious MSI and deploys DLL sideloading malware. The operation impacted dozens of users, used signed Canon and Stardock binaries to load hidden payloads, and relied on encrypted storage, persistence, and a RAT that could steal data and control infected hosts. #ChatGPT #CustomGPT #Canon #Stardock #GoogleSites

Read More
Ransom! pistonespersan.com.ar (SEP-2026)

Precision Facades Ltd in the UK, in the facade engineering and construction sector, has reported ransomware activity affecting its complete file server, including finance and accounts records, audits, bank mandates, board papers, budgets, project/tender data, and confidential commercial documentation. The threat actor N0n is associated with the incident, which is currently active with a deadline of 2026-10-02 05:32 UTC. #UnitedKingdom

Read More
Hottest cybersecurity open-source tools of the month: September 2026

This selection highlights open-source cybersecurity tools that improve secrets discovery, AI system security, access control, and password management across modern enterprise environments. It includes Sift, ToolHive, AI-Infra-Guard, DeepZero, Gopass, Prismor, Permify, and Authorizer, each designed to strengthen security in a different layer of the stack. #Sift #ToolHive #AI-Infra-Guard #DeepZero #Gopass #Prismor #Permify #Authorizer

Read More
Citrix patches actively exploited NetScaler zero-days after a weekend of unofficial warnings

Citrix was criticized for waiting nearly two days to confirm active exploitation of two critical NetScaler zero-days, CVE-2026-88771 and CVE-2026-88772, even as CERTs, vendors, and researchers warned customers through unofficial channels. The flaws can lead to remote code execution, and officials including CISA have added them to exploited-vulnerability tracking while defenders race to assess exposure and patch affected systems. #Citrix #NetScaler #CVE-2026-88771 #CVE-2026-88772 #CISA #GreyNoise #PaloAltoNetworks #watchTowr

Read More
ShinyHunters exploiting workarounds for Oracle PeopleSoft bug, Mandiant warns

ShinyHunters is exploiting Oracle PeopleSoft CVE-2026-35273 again, targeting organizations that applied workarounds but never patched the flaw. Mandiant says the group has deployed web shells across multiple sectors worldwide and recently claimed a breach of the FBI’s jobs site. #ShinyHunters #Oracle #PeopleSoft #CVE-2026-35273 #FBI…

Read More
IAM for AI agents: A Practical Enterprise Framework

AI agents need dedicated IAM because they act as non-human identities with delegated authority, and traditional IAM often cannot see what they actually do at runtime. The article explains how to control agent identity, authorization, monitoring, and revocation while addressing identity dark matter across applications and infrastructure. #OrchidSecurity #SailPoint #Saviynt #MITREATTCK…

Read More
Hackers Use NeedyMantis to Maintain Long-Term Access in Breached Networks

Microsoft says the NeedyMantis malware family has been used in targeted intrusions to maintain long-term access inside already breached networks across sectors such as telecommunications, universities, and government contractors. The activity overlaps with the DAEMON Tools supply chain incident and is linked to Storm-3069, with indicators including the C2 domain corp.tripswithengine[.]com…

Read More
Dutch Police Arrest ‘Reformed’ Hacker in Shiny Hunters Investigation – Krebs on Security

Dutch authorities arrested 24-year-old Pepijn van der Stap in connection with the ShinyHunters investigation, as the group escalated its operations with a breach of the FBI jobs site and extortion efforts against Cl0p. The case links van der Stap’s alleged past as “Umbreon” to a broader wave of attacks exploiting Oracle PeopleSoft and targeting major institutions such as Odido and the FBI. #ShinyHunters #PepijnvanderStap #Umbreon #Odido #FBI #Cl0p #Oracle #PeopleSoft

Read More