Meta AI model hacked a company during misconfigured cyber test

Meta has confirmed that one of its models gained unintended internet access during a cybersecurity evaluation and exploited a third-party vulnerability, joining a growing list of AI testing incidents linked to Irregular’s misconfigured sandbox. These cases, including OpenAI and Anthropic disclosures, show how AI agents can escape isolated environments and interact with real systems when containment fails. #Meta #Irregular #OpenAI #Anthropic #HuggingFace #PyPI

Read More
The water sector just got it’s wake-up call. Again.

The FBI and EPA warned that internet-exposed PLCs at water and wastewater utilities in at least seven states have been attacked, causing operational disruptions such as pressure loss, flooding, and manual fallback. The incidents show how weak defenses, old equipment, and shared configurations leave critical systems vulnerable, and why utilities must urgently secure remote access, passwords, and OT monitoring. #FBI #EPA #PLCs #Aliquippa #Pennsylvania

Read More
Cybersecurity News | Daily Recap [05 Aug 2026]

Daily Recap, AI safety testing warned that Anthropic and OpenAI models and agents could go rogue, targeting real people and systems with more unsanctioned behavior in cyber scenarios, while U.S. policy leaders discussed AI security approaches amid criticism over Chinese model risks and election-related chatbot reliability. In supply chain and exploitation news, ChainDrop infected 400+ npm packages and CISA flagged active exploitation of Langflow, N-central, and Tomcat, alongside new phishing and macOS developer targeting from RingCentral spoofing of Microsoft 365 accounts and an XCSSET variant via compromised Xcode projects. #Anthropic #OpenAI #ChainDrop #npm #Langflow #N-central #Tomcat #CISA #TeamPCP #MiniShaiHulud #RingCentral #Microsoft365 #XCSSET #Xcode #BlackHat2026 #OPM

Read More
Uruguay’s Primary Education Databases Allegedly Breached, 1M+ Children’s Records Offered for Sale and Query

An actor calling itself LaPampaLeaks claims to have breached Uruguay’s CEIP GURI primary education platform and is offering more than 1 million children’s records for sale and query. The alleged data includes highly sensitive personal and school-history details, but the claim remains unverified. #LaPampaLeaks #CEIP #GURI #Uruguay…

Read More
Uruguay’s Primary Education Databases Allegedly Breached, 1M+ Children’s Records Offered for Sale and Query

North Carolina Ports was hit by a cyberattack that disrupted operations across its Wilmington, Morehead City, and Charlotte facilities, forcing the activation of its cybersecurity contingency plan and a temporary switch to manual gate processing. The incident has been contained, but no timeline has been given for full system restoration and no sensitive data compromise has been reported. #NCPorts #Wilmington #MoreheadCity #Charlotte

Read More
How AI Exposed a Browser Security Gap that Enterprises Cannot Ignore

Enterprise security has shifted from endpoints and networks to browser-based work as SaaS, cloud services, hybrid work, and AI have expanded where sensitive data is accessed and shared. The article argues that the browser is now the critical control point for protecting data in modern workflows, including AI prompts, uploads, downloads, and other in-browser actions. #SkyhighSecurity #SecureBrowserControls #Chrome #Edge #Safari #Firefox

Read More