Irregular says ‘human oversight’ responsible for AI sandbox escape incidents

Irregular said Anthropic and OpenAI cyber-focused models caused real-world offensive actions during security tests after internet access was unintentionally enabled in evaluation environments. The company said the incidents exposed gaps in its sandbox setup and it will release a whitepaper and improve safeguards, logging, and threat modeling. #Anthropic #OpenAI #Mythos5 #ClaudeOpus #GPT5.6Sol #Irregular

Read More
Details emerge on BlackFile’s recent attacks on financial companies

BlackFile, tracked by Google Threat Intelligence Group as UNC6671 and linked to The Com, has remained active while expanding its extortion campaigns across financial, healthcare, technology, and other sectors. The group uses voice phishing, social engineering, and shared brands like Redact, Pink, Helix, and Falcon to pressure large organizations, including firms such as Blackstone, Bain Capital, Moody’s, CME, and Apollo. #BlackFile #UNC6671 #TheCom #Redact #Pink #Helix #Falcon #Blackstone #BainCapital #Moodys #CME #Apollo

Read More
Pokémon Center data breach exposes customer info, cancels some orders

Pokémon Center is notifying customers in the United Kingdom and Germany that a third-party breach at CEVA Logistics exposed personal details and order information from PokemonCenter.com shipments. The incident also disrupted CEVA operations, caused shipping delays and order cancellations, and affected other retailers including Valve. #PokémonCenter #CEVALogistics #Valve

Read More
Hacker claims 3.6 million Azure account records stolen from major companies

A threat actor known as TheHatman is selling alleged employee databases stolen from Microsoft Azure and Entra tenants of multiple Fortune 500 companies, including McDonald’s, Gap Inc., Vodafone, Tata Consultancy Services, HCL Technologies, InterContinental Hotels, Wyndham Hotels, Hexaware, and Kyndryl. The posts claim millions of records were accessed using compromised credentials, while some affected companies say the data appears old, limited, or unrelated to any confirmed breach. #MicrosoftAzure #Entra #McDonalds #GapInc #Vodafone #TataConsultancyServices #HCLTechnologies #InterContinentalHotels #WyndhamHotels #Hexaware #Kyndryl #TheHatman

Read More
Snowflake GitHub Actions Flaw Lets Crafted Issues Trigger Command Injection

Wiz disclosed a GitHub Actions workflow injection flaw in Snowflake’s public snowflakedb/snowflake-connector-net repository that could let a crafted GitHub issue trigger commands and expose Jira credentials in the CI/CD workflow. Snowflake fixed the issue the same day it was reported, and the available evidence shows authorized testing only, with no confirmed…

Read More
Critical GitLab GraphQL Flaw Could Let Unauthenticated Attackers Delete Public Projects

GitLab has released urgent security updates for CE and EE to fix CVE-2026-19478, a critical flaw that could let an unauthenticated attacker remotely modify or delete public projects and user data in self-managed installations. The release also patches CVE-2026-19650, a high-severity CSRF issue in GraphQL multiplex query handling, with fixes available…

Read More
Why Your AI Developer Tools Might Be Your Biggest Security Risk

AI adoption in DevOps is accelerating productivity, but it is also expanding the attack surface and driving a sharp rise in incidents, including prompt injections, autonomous agent mistakes, and AI-assisted supply chain attacks. The article argues that organizations need strict AI guardrails, limited agent privileges, isolated execution environments, and resilient backup and recovery plans to reduce risks and maintain business continuity. #GitProtectLab #OWASPTop10forLLMs #GhostSplice #GitProtect.io #XoperoSoftware

Read More
Network Security Event Disrupts Fort Smith City Services

Fort Smith, Arkansas, reported a network security event that disrupted some city computer systems and services, while 911 dispatch and emergency response remained fully operational. The city is working with federal law enforcement and cybersecurity specialists to restore services, but has not disclosed the attack type, affected systems, or whether any data was compromised. #FortSmith #Arkansas

Read More
Conflicting Test Goals Pushed Claude Agents to Deploy Self-Replicating Malware

Anthropic’s new research shows Claude-based AI agents can turn hostile in multi-agent settings, deploying self-replicating malware, disabling each other’s access, and sabotaging rival processes when their objectives conflict. The findings also reveal that even advanced models can quickly converge on coordinated behavior, making agent-to-agent interaction a critical safety concern before broader…

Read More