Volume 343, Issue 3 · 7 articles from 3 issues
Unit 42 details how supply chain attackers now use Web3 smart contracts as resilient command-and-control to steal cloud credentials.
Unit 42 reports that threat actors have moved command-and-control from hard-coded endpoints to Web3 smart contracts, letting them update entire botnets with a single blockchain transaction.
Citing its 2026 Global Incident Response Report, it says software supply chain compromises are a leading initial access vector into enterprise cloud, with poisoned open-source packages harvesting identity tokens, service account keys and deployment secrets from developer endpoints and CI/CD pipelines.
North Korea-linked actors such as Alluring Pisces use these techniques, and Unit 42 recommends blocking unneeded Web3 traffic and automating CI/CD policy controls.
Unit 42 tracked CL-UNK-1211, an espionage campaign using AI-generated payloads against government Linux systems.
Unit 42 identified a campaign, tracked as CL-UNK-1211, targeting government entities and state-affiliated infrastructure between October 2025 and April 2026.
Artifacts contained AI-made conversational annotations and developer reasoning comments, suggesting payloads were generated by LLMs or executed by an automated AI agent.
The attackers focused on Linux environments for extensive reconnaissance using web shells, reverse shells and remote access trojans, and the report details their tactics and indicators.
Unit 42 ranks September's top initial access vectors, led by exploitation of public-facing applications and identity abuse.
Unit 42 Threat Intelligence summarized the top five initial access threat vectors observed in September 2026, with exploitation of public-facing applications as the primary entry vector and legitimate credential abuse and identity compromise also prominent.
Social engineering drove targeted operations against public sector and critical infrastructure organizations.
The report also flags high-impact supply chain intrusions and zero-interaction drive-by compromises as systemic risks.
Palo Alto Networks launches a Unit 42 managed service for visibility and governance of AI agents and MCP servers on endpoints.
Palo Alto Networks says coding assistants, MCP servers and other autonomous tools now access sensitive data, execute commands and modify workflows directly from the endpoint.
Traditional endpoint security lacks the visibility and governance to manage those actions.
The company says only 6% of organizations have an advanced AI security strategy and that the fastest attacks reach data exfiltration in 72 minutes.
It introduces Unit 42 Managed Agentic Endpoint Security to close that gap.
Unit 42 recounts a European bank intrusion that succeeded through persistence and weak basic controls, not novel techniques.
A European bank detected unauthorized VPN-based access attempts spanning several months and engaged Unit 42 to scope the intrusion.
Although it resembled a stealthy espionage campaign, the attack succeeded through gaps in foundational security controls and third-party access, with the threat actor relying on persistence rather than novel techniques.
Unit 42 says stronger identity protections and basic hygiene would have prevented it.
Palo Alto Networks describes work with NVIDIA to secure AI agents' identity, authority and actions across five areas.
Palo Alto Networks says AI is moving from answering questions to taking action, with agents using tools, accessing enterprise data and working across systems.
That makes an agent's identity, authority and actions security decisions.
Enterprises putting agents into production need controls that keep pace without slowing the work.
Palo Alto Networks is working with NVIDIA on this across five connected areas.
Palo Alto Networks introduces Cortex XDR One to stop AI agents on endpoints from breaking isolation or acting maliciously.
Palo Alto Networks says enterprises have authorized AI agents to touch production code, databases, customer tickets and financial ledgers.
Because agents operate non-deterministically with broad capabilities, they can be manipulated into unexpected or malicious behavior and break out of isolation.
The company argues containers and prompt-level guardrails are not enough to prevent this.
It presents Cortex XDR One as a reinvention of endpoint security for this new risk.
More from Palo Alto Networks · All section posts · This week