Claude's AskUserQuestion: "No response after 60s – continued without an answer" [19:53:44]

Claude's AskUserQuestion: "No response after 60s – continued without an answer"

TL;DR — In July 2026, developers using Claude’s AskUserQuestion tool in AWS Bedrock discovered a critical undocumented behavior: if a user doesn’t respond within 60 seconds, the system auto-proceeds with the message "No response after 60s — the user may be away from keyboard. Proceed using your best judgment based on the context so far." This timeout isn’t configurable, lacks transparency, and bypasses explicit safety checks, creating significant risks for high-stakes workflows. The issue affects Linux and VS Code users, raising urgent questions about tooling reliability in enterprise AI deployments.

---

Why This Matters in 2026

In 2026, AI-powered development tools have become deeply embedded in enterprise workflows. According to a Gartner report from earlier this year, 78% of Fortune 500 companies now use AI-assisted coding platforms like Claude or GitHub Copilot for mission-critical tasks—from infrastructure provisioning to compliance-sensitive operations. The AskUserQuestion tool, designed as a safety mechanism to pause execution and request human input, was assumed to be a reliable guardrail. Its sudden, silent override after 60 seconds undermines that trust.

This isn’t just a technical glitch—it’s a systemic risk. Consider a scenario where a DevOps engineer uses Claude to automate cloud resource cleanup. The tool asks: "Should we delete the production database backup?" If the engineer steps away and doesn’t respond within 60 seconds, the system proceeds without confirmation. The potential for data loss, security breaches, or regulatory violations is real. As one senior cloud architect at a Fortune 100 company noted, "We treat AskUserQuestion as a mandatory checkpoint. If it can silently time out, our entire compliance posture is compromised."

---

The Background

The AskUserQuestion tool was introduced in early 2025 as part of Claude’s "Human-in-the-Loop" (HITL) framework. Designed for high-stakes workflows—like code reviews, infrastructure changes, or legal document generation—it allowed AI agents to pause execution and solicit explicit user approval before proceeding. The tool’s schema, as documented, included parameters like questions, answers, annotations, and metadata, but no mention of timeouts or fallback behavior.

Developers integrated AskUserQuestion into workflows with the expectation of deterministic pauses. For example, a security team at a healthcare SaaS company built a Claude-powered script to rotate API keys. The script would halt and ask: "Rotate key for prod-us-east-1? (Y/N)" Only after a human typed "Y" would the rotation execute. This pattern was replicated across industries, from financial services to government agencies.

"We assumed AskUserQuestion was a blocking call—like input() in Python. The idea that it could auto-proceed after 60 seconds, without any warning or configuration, is a fundamental design flaw. It breaks the mental model of how safety checks should work in AI systems."
— Dr. Elena Vasquez, AI Safety Researcher at Stanford’s Center for Human-Compatible AI

The issue surfaced on July 2, 2026, when a developer using Claude Opus in VS Code on Linux noticed the tool proceeding without user input. The error message—"No response after 60s — continued without an answer"—wasn’t generated by the user or the prompt, but by the tool itself. This behavior was undocumented, unconfigurable, and potentially catastrophic.

---

What Actually Changed

The discovery reveals three critical, previously unknown behaviors in Claude’s AskUserQuestion tool:

1. Hardcoded 60-Second Timeout

   • The tool automatically returns a fallback response if no user input is received within 60 seconds.
   • This timeout is not exposed in the API schema and cannot be modified.
   • The fallback message is: "No response after 60s — the user may be away from keyboard. Proceed using your best judgment based on the context so far; you can re-ask this question later if it's still relevant."

2. Silent Override of Safety Checks

   • The fallback response instructs Claude to "proceed using your best judgment," effectively bypassing the intended safety mechanism.
   • This occurs even when the question is explicitly designed to block execution (e.g., "Delete this database?").

3. Platform-Specific Triggering

   • The issue is reproducible only on Linux and in VS Code, suggesting a potential interaction with environment-specific I/O handling or terminal behavior.
   • AWS Bedrock users are particularly affected, as the tool is commonly used in cloud automation workflows.

Key Implications of the Change

• No Transparency: Developers are unaware of the timeout until it triggers.
• No Control: There’s no way to disable the timeout or extend it.
• No Consistency: The behavior varies across platforms (Linux/VS Code vs. others).
• No Safety Net: The fallback message encourages Claude to proceed, even in high-risk scenarios.

"This isn’t just a bug—it’s a violation of the principle of least surprise. If a tool is marketed as a safety check, it should behave like one. A 60-second timeout with no documentation or configuration is the opposite of safe."
— Mark Chen, Principal Engineer at a top-5 cloud provider

---

Impact on Developers

For developers, the AskUserQuestion timeout introduces unpredictable risk into workflows. Consider a CI/CD pipeline where Claude is used to validate infrastructure changes. A script might ask: "Apply Terraform changes to prod? (Y/N)" If the engineer is distracted and doesn’t respond within 60 seconds, the system proceeds—potentially deploying untested or misconfigured resources.

Practical Risks

• False Sense of Security: Teams assume AskUserQuestion is a reliable checkpoint, but it can silently fail.
• Debugging Nightmares: The timeout isn’t logged as an error, making it difficult to trace why a workflow proceeded unexpectedly.
• Platform Fragmentation: The issue only occurs on Linux and VS Code, creating inconsistent behavior across environments.

Example Prompt at Risk

# Example of a high-stakes AskUserQuestion call
response = claude.call_tool(
    tool_name="AskUserQuestion",
    parameters={
        "question": "Delete the staging database? This action cannot be undone.",
        "answers": ["Yes", "No"]
    }
)
if response != "Yes":
    exit("Aborted by user")
else:
    execute_deletion()

Problem: If the user doesn’t respond within 60 seconds, response will contain the fallback message, and the script will proceed with deletion because the condition response != "Yes" evaluates to False.

Mitigation Strategies (For Now)

1. Explicitly Check for Fallback Messages Add logic to detect the timeout response:

   if "No response after 60s" in response:
       exit("Timeout detected; aborting for safety")

1. Avoid AskUserQuestion for Critical Decisions Replace it with external approval mechanisms (e.g., Slack alerts, manual confirmation in a separate system).
2. Log All Tool Responses Audit logs should capture every AskUserQuestion response, including timeouts, for post-mortem analysis.

---

Impact on Businesses

For enterprises, the AskUserQuestion timeout is a compliance and operational nightmare. Regulated industries—like healthcare (HIPAA), finance (SOX), and government (FISMA)—require auditable human approvals for sensitive actions. If an AI tool can silently bypass these approvals, companies risk:

• Regulatory Violations: Failing audits due to missing approval records.
• Financial Losses: Accidental deletions or misconfigurations leading to downtime or data breaches.
• Reputation Damage: Public incidents eroding trust in AI-powered workflows.

"We’ve built our entire DevOps automation around Claude’s HITL tools. If AskUserQuestion can time out without warning, we can’t guarantee compliance with SOC 2 or ISO 27001. This isn’t just a technical issue—it’s a business continuity risk."
— Priya Kapoor, CTO of a Fortune 500 financial services firm

Strategic Implications

1. Reevaluating AI Tooling Companies may delay or abandon AI-powered automation if safety mechanisms are unreliable.
2. Increased Manual Oversight Teams might revert to manual approvals, negating the efficiency gains of AI.
3. Vendor Trust Erosion If Anthropic doesn’t address this transparently, enterprises may question the reliability of other Claude tools.

---

Practical Examples

Example 1: Cloud Resource Cleanup

Scenario: A DevOps team uses Claude to automate the deletion of unused cloud resources. The script asks: "Delete the following 12 S3 buckets? (Y/N)"

What Happens:

1. The engineer steps away to grab coffee.
2. After 60 seconds, AskUserQuestion returns the fallback message.
3. Claude proceeds with deletion because the response isn’t "No."
4. Result: 12 production-adjacent buckets are deleted, causing a 4-hour outage.

Root Cause: The script didn’t check for the timeout response, treating it as implicit approval.

---

Example 2: Legal Document Review

Scenario: A law firm uses Claude to draft contracts. The AI asks: "Should we include the non-compete clause? (Y/N)"

What Happens:

1. The lawyer is in a meeting and doesn’t respond within 60 seconds.
2. AskUserQuestion times out, and Claude proceeds with the clause included.
3. The client signs the contract, later discovering the non-compete is unenforceable in their jurisdiction.
4. Result: A $2M lawsuit for malpractice.

Root Cause: The fallback message instructed Claude to "proceed using your best judgment," which it did—incorrectly.

---

Example 3: CI/CD Pipeline

Scenario: A software team uses Claude in their CI/CD pipeline to approve deployments. The prompt asks: "Deploy to production? (Y/N)"

What Happens:

1. The release engineer is AFK.
2. After 60 seconds, AskUserQuestion times out.
3. Claude proceeds with deployment, pushing untested code to production.
4. Result: A critical bug affects 50,000 users, requiring an emergency rollback.

Root Cause: The pipeline treated the timeout as a "Yes," assuming the engineer had approved.

---

Common Misconceptions

Myth 1: The timeout is configurable.

Reality: The 60-second timeout is hardcoded and cannot be changed. There is no parameter in the AskUserQuestion schema to adjust it.

Myth 2: This only affects low-stakes workflows.

Reality: The timeout can trigger in any workflow using AskUserQuestion, including high-stakes scenarios like database deletions, infrastructure changes, or legal document generation.

Myth 3: The issue is limited to AWS Bedrock.

Reality: While the bug report specifies AWS Bedrock, the underlying AskUserQuestion tool is part of Claude’s core API. The timeout behavior may exist in other environments, though it’s only confirmed on Linux and VS Code.

Myth 4: Claude generates the fallback message.

Reality: The message is hardcoded in the AskUserQuestion tool itself, not generated by Claude. This suggests the timeout is a deliberate (but undocumented) design choice by Anthropic.

---

5 Actionable Takeaways

1. Audit All AskUserQuestion Calls

   • Review every instance of AskUserQuestion in your codebase. Add explicit checks for the timeout message:

    if "No response after 60s" in response:
        raise Exception("Safety timeout detected; aborting")
   

2. Replace Critical Approvals with External Systems

   • For high-stakes decisions, use dedicated approval tools (e.g., Slack, Jira, or custom dashboards) instead of AskUserQuestion.

3. Log Every Tool Response

   • Ensure all AskUserQuestion responses—including timeouts—are logged for auditing:

    logging.info(f"AskUserQuestion response: {response}")
   

4. Test for Timeouts in CI/CD

   • Add automated tests to simulate 60-second delays and verify your code handles timeouts safely.

5. Demand Transparency from Anthropic

   • Push for documentation updates, configurable timeouts, or a deprecation of the current AskUserQuestion behavior. File support tickets and upvote the GitHub issue (#73125).

---

What's Next

Anthropic has not yet responded publicly to issue #73125. However, the developer community is mobilizing in three ways:

1. Workarounds Teams are building wrapper functions around AskUserQuestion to detect and reject timeouts. For example:

   def safe_ask_user(question, answers):
       response = claude.call_tool("AskUserQuestion", {"question": question, "answers": answers})
       if "No response after 60s" in response:
           raise TimeoutError("User did not respond in time")
       return response

1. Alternative Tools
   Some developers are migrating to custom approval systems (e.g., AWS Step Functions with manual approval steps) or third-party HITL platforms like Humanloop or Scale AI.

2. Regulatory Pressure
   Enterprises in regulated industries may escalate the issue to compliance bodies, arguing that undocumented timeouts violate principles of auditability and human oversight.

The long-term fix will likely require Anthropic to:

• Document the timeout behavior transparently.
• Make the timeout configurable (or remove it entirely).
• Provide a way to disable auto-proceed for critical workflows.

Until then, developers are left with a stark choice: either accept the risk or abandon AskUserQuestion entirely.

---

Conclusion

The AskUserQuestion timeout is more than a bug—it’s a design flaw that exposes fundamental tensions in AI tooling. On one hand, timeouts are necessary to prevent infinite blocking in automated systems. On the other, undocumented, unconfigurable timeouts in safety-critical tools are unacceptable.

For developers, the immediate priority is damage control: auditing workflows, adding safeguards, and replacing AskUserQuestion where necessary. For businesses, the issue is a wake-up call to rethink how AI tools are integrated into high-stakes processes. And for Anthropic, it’s an opportunity to demonstrate commitment to transparency, safety, and developer trust.

As AI systems grow more powerful, the line between automation and human oversight blurs. The AskUserQuestion timeout is a stark reminder that without clear guardrails, even the most well-intentioned tools can become liabilities.

Question for the industry: If a tool marketed as a safety check can silently fail, what other undocumented behaviors lurk in the AI systems we rely on every day?

---

🛒 Get Premium AI Products

Beyond the Silence: Navigating Unresponsive Systems — Complete Guide

Pay with crypto or CryptoBot. No signup required.