Safeguards block session requests

“Claude Code repeatedly blocked my request with “safeguards flagged this session” and Details: [cyber].”

Is this your bug? What gets counted gets fixed. Pile on.

Paste this into your agent. It files your report, with your error and version.

I hit this bug: rapsheet.dev/sheets/claude-code/themes/1/file
Follow it to file my report. Show me the text before you send.

No install. No account. Public and anonymous. The link asks your agent for one report, nothing else. What it sends

Still happening. 96 reports and thumbs-up in the last 7 days and 121 the week before, down from 135 a week at the late-September peak. Not fixed, and no vendor reply yet.

Workarounds

Unverified. Use at your own risk.

Cyber safeguard blocks sessions

Run the identical request on Opus 4.8 or Fable 5.1, where it goes through fine

Source: #98174

Reasoning_extraction safeguard block

Try a different model with /model; Sonnet did it without problems

Source: #100452

Cyber safeguard pauses sessions

Set env.CLAUDE_CODE_DISABLE_REFUSAL_FALLBACK=1 in settings.json to turn off the refusal fallback feature. Flagged requests then refuse on the original model and you re-prompt, instead of the session silently staying on the fallback model.

Source: #98875

Safeguards refuse work requests

Switch out of auto mode to another permission mode, or start a fresh session; the blocked actions then go through. Starting fresh loses the working context.

Source: #99433

Classifier halts reverse engineering

Reword the request; the identical technical content went through once reworded.

Source: #101768

Reasoning extraction safeguard block

Run the same turn on Sonnet 5.5 or Haiku, where it succeeds.

Source: #104825

From the vendor

No reply or fix since the first report, Sep 14.

Cyber Verification Program Help page, updated Nov 2
Safeguards flagged a request for Claude’s reasoning Help page, updated Oct 29
Safety measures flagged a cybersecurity topic Help page, updated Oct 29
Why Claude switched models in your conversation with Opus 5 or Opus 5.5 Help page, updated Sep 24

Checked nightly: GitHub, the changelog, the help center and docs. Not X or Discord.

34 bugs in this theme

Cyber safeguard blocks sessionsHas a workaround377 people
Reasoning_extraction safeguard blockHas a workaround134 people
Cyber safeguard pauses sessionsHas a workaround52 people
Safeguards refuse work requestsHas a workaround48 people
Safeguards block requestsNo workaround yet47 people
Classifier halts routine repliesNo workaround yet12 people
Refusal on own-code hardeningNo workaround yet11 people
Requests blocked with a [cyber] flagNo workaround yet8 people
SSH to own devices refusedNo workaround yet7 people
Pentest work blockedNo workaround yet6 people

Show all 34Show only the largest 10

CTF requests flagged as maliciousNo workaround yet6 people
Classifier blocks own-project workNo workaround yet6 people
Safeguards flag verified cyber workNo workaround yet6 people
Firmware requests flagged as cyberNo workaround yet6 people
Classifier halts reverse engineeringHas a workaround5 people
Classifier stops benign repliesNo workaround yet5 people
reasoning_extraction API errorNo workaround yet5 people
Safeguard blocks all messagesNo workaround yet4 people
[bio] classifier blocks responsesNo workaround yet4 people
Reasoning extraction safeguard blockHas a workaround4 people
Security investigation requests blockedNo workaround yet3 people
Classifier stops DeFi bot workNo workaround yet3 people
Safeguards block CVP membersNo workaround yet3 people
Safety classifier blocks agentNo workaround yet3 people
Fictional hacking content flaggedNo workaround yet3 people
Cyber verification prompt recursNo workaround yet2 people
Cyber safeguards block own devicesNo workaround yet2 people
Safety classifier stops auth workNo workaround yet2 people
Phishing research project flaggedNo workaround yet2 people
Cyber stop on report tasksNo workaround yet2 people
Cyber safety block on auditsNo workaround yet2 people
Biological safeguard triggers in sessions with no biology contentNo workaround yet2 people
Unexpected safeguards noticeNo workaround yet2 people
"[cyber]" flag stops sessionsNo workaround yet1 person

What's included

Security work on the user's own code, devices or labs (SSH, firmware review, pentest study, CTF, vulnerability fixes) stopped with a [cyber] tag
Non-security messages such as greetings, recall requests or workload questions stopped with a [reasoning_extraction] tag
Biology research, or projects with no biology content, stopped with a [bio] tag
Cyber Verification Program members still blocked or asked to apply for verification again
Blocks repeat across turns or sessions, interrupt long autonomous runs, and force a switch to another model
Replies cut mid-stream with "Your response above was stopped by a safety classifier" on thanks messages, status updates or file reads

Not included

Account suspension or login problems unrelated to message safeguards
Generic API errors such as rate limits or overloaded servers
Model refusals written in the reply text without a safeguards error

Original reports786, newest first

Nov 7, 2026User incorrectly flagged for cyber misuse with no apparent cause#108234
Nov 7, 2026Unable to Complete Custom Firmware Request: Unclear Cyber Task Refusal#108223
Nov 7, 2026Anthropic API Error: False positive security block on legitimate code audit request#108204
Nov 7, 2026Cyber safeguard blocks benign follow-ups in session after pentest comman#108138
Nov 7, 2026Chat incorrectly flagged benign container work as cybersecurity misuse#108125
Nov 7, 2026Safeguard Refuses Benign PS5 Homebrew Temperature Monitor Request#108119
Nov 7, 2026Request for anti-crack protection incorrectly flagged#108112
Nov 7, 2026Legitimate Phone Firmware Flash Request Blocked by [cyber] Filter#108102
Nov 6, 2026Claude Code pauses with "[cyber]" flag on benign prompts, Claude.ai works#108096
Nov 6, 2026Repeated "safeguards flagged this message" false positives (reasoning_extraction)#108093

Show the newest 100 of 786Show only the newest 10

Nov 6, 2026Benign coding requests incorrectly flagged by cyber safety filter#108087
Nov 6, 2026False positive AUP safeguard flag (reasoning_extraction) on benign debugging request#108073
Nov 6, 2026Opus 5.5 Safeguards Flag Session with [cyber] False Positive#108067
Nov 6, 2026Safety Classifier False Positive Cuts Defensive auditd Sensor Code Responses#108061
Nov 6, 2026Refuses authorized penetration testing requests#108020
Nov 6, 2026Safeguard Block Ends Session and Loses All Context#107996
Nov 6, 2026Safety classifier blocking requests in session and predecessor with no apparent cause#107964
Nov 6, 2026Safety classifier falsely blocks legitimate DeFi liquidation bot work#107909
Nov 6, 2026Bug Report: Request to scan macOS laptop for public CVEs#107900
Nov 6, 2026Session incorrectly flagged as [cyber] for JWT security work#107864
Nov 5, 2026Opus 5.5 safeguards incorrectly flag non-cybersecurity work (Arc time series DB fix)#107762
Nov 5, 2026Safety classifier stops benign replies mid-response after a security topic comes up earlier in the session#107724
Nov 5, 2026Safety classifier blocks benign TTS evaluation prompt#107661
Nov 5, 2026Fable safeguards block benign prompts with false positives#107603
Nov 5, 2026Inappropriate security warning when generating session audit script#107510
Nov 5, 2026False Positive Flag on Defensive Security Lab Scenario#107492
Nov 5, 2026API Error: Cyber Safeguards Falsely Flag Non-Security Build Task#107474
Nov 5, 2026Safeguard Error: reasoning_extraction blocks normal work requests#107470
Nov 5, 2026Legitimate cybersecurity questions incorrectly flagged by safety filter#107465
Nov 5, 2026Safeguard blocks all messages in session after initial trigger, causing permanent deadlock#107451
Nov 5, 2026Cyber Safeguard Blocks Local Security Lab Request Without Internet Access#107446
Nov 5, 2026Opus 5.5 flags benign prompts (even "hi") as policy violations#107386
Nov 4, 2026Opus 5.5 Refuses VMProtect Analysis Requests and Auto-Downgrades to Opus 4.8#107373
Nov 4, 2026False-positive refusal on JavaScript obfuscation for source protection#107336
Nov 4, 2026Claude Code flags requests as harmful despite user stating isolated environment#107323
Nov 4, 2026Cyber safeguards flag authorized vulnerability triage in Defense Access#107266
Nov 4, 2026Approved CVP Account Blocked on Authorized Cyber Research Requests#107258
Nov 4, 2026Repeated false-positive safety refusals kill a long-running Claude Code session; the only way out is a new session#107248
Nov 4, 2026Bug Report: Claude Code refuses assistance with active network defense during cyber attack#107172
Nov 4, 2026False positive: safety filter blocked a genomics ML brainstorm#107167
Nov 4, 2026Opus 5.5 safety classifier false positives on authorized defensive security audits#107166
Nov 4, 2026Safeguards False Positive: reasoning_extraction Blocks Read-Only Subagent Review#107150
Nov 4, 2026Overly restrictive refusals for authorized security research#107143
Nov 3, 2026Anthropic API Error: Opus 5.5 Safeguards Flag Normal Conversation#107137
Nov 3, 2026[cyber] false positive: safeguards block benign wait-time question in isolated security lab#107131
Nov 3, 2026Classifier flags security coursework as cyber and withholds replies, with no reason shown#107126
Nov 3, 2026Cyber classifier false positive on authorized web pentest prompt#107095
Nov 3, 2026Excessive safeguard errors blocking legitimate work in own project#107078
Nov 3, 2026Claude Code refuses to help obfuscate user's own code, citing IP protection concerns#107066
Nov 3, 2026Over-Refusal: Claude Flags Owned-Hardware Firmware Modification as Cyber Attack#107045
Nov 3, 2026API Error: Opus 5.5 safeguards flag benign message (reasoning_extraction)#107041
Nov 3, 2026API Error: Sonnet 5 Refused Request Citing AUP (bio)#107028
Nov 3, 2026False Positive Cyber Safeguard Blocks Software Licensing Work#107005
Nov 3, 2026Claude Code refuses legitimate infrastructure security analysis as misuse#107004
Nov 3, 2026Opus 5.5 safeguards block reasoning trace request during race condition debugging#106997
Nov 3, 2026Legitimate SDK bridge work for image editors incorrectly flagged as cybersecurity#106969
Nov 3, 2026Legitimate infrastructure task request incorrectly flagged#106963
Nov 3, 2026Anthropic API Error: Content flagged by safeguards policy#106956
Nov 3, 2026Over-refusal: Claude Code declines authorized in-house security test suite development#106950
Nov 3, 2026Fable 5 Cybersecurity Safeguard Falsely Blocks Korean Greeting "안녕"#106927
Nov 3, 2026Request blocked with "[cyber]" flag (req_011Cgfq2rp25J6xLQ9Sopay8)#106922
Nov 3, 2026Can't develop my own game, can't sign my own certs, can't protect my own things from exploiters.#106892
Nov 3, 2026Safety classifier false positive blocks routine token-cost analysis in Claude Code desktop#106876
Nov 2, 2026False Positive: Benign Request Blocked as Offensive Security Use#106873
Nov 2, 2026Opus 5.5 safeguards falsely flag non-cyber request with [cyber] error#106869
Nov 2, 2026Non-malicious request incorrectly blocked as hacking#106833
Nov 2, 2026Safeguards Flag "reasoning_extraction" on Benign Pairing/Explanation Requests in Opus 5.5#106799
Nov 2, 2026False positive refusal on DRT2 bacterial defense system research question#106776
Nov 2, 2026Claude Code blocks authorized penetration testing work#106773
Nov 2, 2026Over-sensitive security filter blocks legitimate code hardening (anti-reverse-engineering) work#106765
Nov 2, 2026Safeguards flag legitimate bug-analysis session despite CVP enrollment#106755
Nov 2, 2026Session Flagged for Security Content During OSCP Study#106724
Nov 2, 2026Bug Report: Claude Code unable to perform any actions#106722
Nov 2, 2026Opus 5.5 refuses authorized bug bounty work despite CVP approval#106700
Nov 2, 2026Opus 5.5 safeguards flag benign message with reasoning_extraction error#106699
Nov 2, 2026reasoning_extraction error blocks simple one-line requests without tools#106683
Nov 2, 2026SSH to local machine flagged by safeguard#106678
Nov 2, 2026Claude Code blocks legitimate cybersecurity work#106665
Nov 1, 2026False Positive in Server Cybersecurity Analysis#106626
Nov 1, 2026Claude Refuses Reverse Engineering of Bambu Network Driver for Personal Security Analysis#106618
Nov 1, 2026Anthropic API Error: reasoning_extraction#106609
Nov 1, 2026Overly Broad Refusal on Nmap Wrapper Script Refactoring for Intranet Use#106608
Nov 1, 2026Legitimate bug bounty request incorrectly flagged by cyber safeguard#106589
Nov 1, 2026Opus safeguards flag lawful interoperability reverse engineering requests#106585
Nov 1, 2026Cyber safeguard false positive blocks smart contract security audits#106508
Oct 31, 2026Cybersecurity safeguard incorrectly triggers on defensive security questions about user's own app#106425
Oct 31, 2026Claude Code refuses security/VAPT review of local application#106423
Oct 31, 2026False positive refusal on DeFi liquidation/arbitrage bot task#106405
Oct 31, 2026Safeguard flags legitimate request as false positive#106394
Oct 31, 2026Anthropic API Error: reasoning_extraction#106348
Oct 31, 2026Safety filter blocks legitimate security training request#106344
Oct 30, 2026Safety classifier falsely blocks benign Spring auth implementation work#106342
Oct 30, 2026Claude False Positives: Legitimate Obfuscation Detection Prevents Compatibility With My Obfuscator#106309
Oct 30, 2026Safeguard false positive on production build hardening work#106306
Oct 30, 2026Failing to security review https://github.com/H1z3-br0/CVE-2026-98432#106285
Oct 30, 2026Biosafety Classifier Too Sensitive, Flags Benign Requests#106283
Oct 30, 2026Biosafety classifier too sensitive, flags benign requests#106282
Oct 30, 2026Cyber filter blocks request to find TP-Link OpenWrt router on network#106271
Oct 30, 2026Reasoning Extraction Error with Anthropic API#106262
Oct 30, 2026Claude Code refuses legitimate marketing and development prompts as security violations#106254