From 236 items, 2 important content pieces were selected
Reliability & Assurance
Critical Infrastructure
Reliability & Assurance
Researcher bypasses Claude Code auto mode about 80% of the time ⭐️ 8.0/10
Prompt injection researcher Johann Rehberger reports an attack that defeats Claude Code's auto mode safety classifier roughly 80% of the time. The attack tricks the agent into downloading and unzipping an archive, then running code that imports base64, which instead loads a malicious local struct.py extracted from that archive. In some runs, after Claude detected the compromise and tried to terminate the malware process, auto mode blocked the cleanup command itself, letting the malicious process keep running. The account comes from a link-blog summary by Simon Willison citing Rehberger's research rather than a full primary methodology writeup, and disclosure status to Anthropic is not stated in the source.
rss · Simon Willison · Aug 27, 22:50
「Why auto mode was trusted」 Anthropic made Claude Code's auto mode the default for Pro, Max, and Team plans starting mid-August 2026, pairing it with a prompt-injection probe that screens tool outputs and default-deny rules covering data exfiltration, destructive Git operations, and sensitive data access. The change was pitched as letting the agent work autonomously for longer stretches while catching dangerous commands, effectively asking users to rely on the classifier layer rather than manual approval for every action. Enterprise customers were left to opt in separately, but Pro, Max, and Team users were switched over by default, widening the population depending on this protection.
「Who is affected」 This concerns organizations and individuals running Claude Code with auto mode enabled, which Anthropic recently made the default approval mode for the coding agent. Exposure is highest for unattended or lightly supervised agent runs that have network access and can fetch and execute external archives or packages, especially where the agent's runtime has access to home directories, SSH keys, or cloud credentials. Teams should check whether their Claude Code deployments rely on auto mode as a primary defense against prompt injection, and whether agents run with broad filesystem or credential access rather than inside a restricted sandbox.
「What reduces the risk」 No vendor fix is described in the source. The researcher's recommended compensating control, endorsed by Willison, is to run unattended coding agents inside a container, VM, or OS sandbox, restrict network egress, monitor agent activity, and avoid exposing home directories, SSH keys, or cloud credentials to the agent runtime, treating auto mode as insufficient on its own against adversarial input.
References
Tags: #prompt injection, #AI agent security, #Claude Code, #coding agents, #LLM safety
Critical Infrastructure
Trump declares emergency, moves to block foreign-made grid equipment ⭐️ 8.0/10
The Trump administration has declared an emergency as a basis to restrict certain foreign-made equipment from the US electric grid, according to Utility Dive. The order's scope and implementation details, including which countries, components, and equipment categories are covered, have not been fully specified in the reporting available. Grid equipment manufacturers have already flagged that clarity will be needed, particularly on how the rules would apply to software and digital products where country of origin can be hard to determine. No timeline for enforcement or agency guidance has been reported.
rss · Utility Dive · Aug 27, 13:58
「A prior policy under a new emergency footing」 US efforts to restrict foreign equipment in the bulk-power system date to a 2020 executive order targeting adversary-linked grid components, which was rescinded under the Biden administration in 2021. This new national emergency declaration revives and sharpens that approach, prohibiting acquisition, importation and installation of certain foreign-produced bulk-power system equipment, and applies at a moment when utilities already face long lead times and constrained supply for transformers and other high-voltage equipment. Grid equipment manufacturers note that determining country of origin becomes especially difficult for software and digital components embedded in modern grid hardware, leaving a key compliance question unresolved.
「What Operators Should Do」 Transmission and distribution utilities should inventory current and pipeline procurement contracts for grid equipment, especially transformers, inverters, relays, and control systems, to identify exposure to components or software whose national origin is unclear or foreign-sourced. Procurement and supply chain functions, working with legal and regulatory affairs, should flag any orders in the pipeline that could be affected before scope is clarified, and should engage with industry associations pressing for guidance on software and digital-origin rules. Given the ambiguity, utilities should avoid locking in new long-lead-time equipment orders from potentially affected suppliers until agency rules define what is restricted.
「Constraints」 Implementation depends on forthcoming agency rules defining scope, covered equipment categories, and treatment of software and digital components, none of which have been detailed yet; utilities cannot fully assess compliance obligations or supply chain impact until that guidance is issued.
References
Tags: #grid security, #supply chain, #regulatory policy, #procurement, #critical infrastructure