Responsible disclosure across jurisdictions
TL;DR: “Responsible disclosure” looks the same in technical writeups but is legally inconsistent across jurisdictions. The US (CFAA), UK (Computer Misuse Act 1990), EU (NIS2 + national implementations), Germany (StGB §202c), Japan, Brazil, and others all treat unauthorised testing differently. Researchers operating internationally must understand: scope of authorisation, safe-harbour legalities, journalist-protection limits, export control on PoC code, and bounty-program contractual protections. Companion to report-writing-step-by-step and disclosure-and-comms.
Why it’s complicated
Security research often:
- Tests against a system that may be in another country.
- Discloses to a vendor headquartered elsewhere.
- Publishes a writeup readable globally.
- May require sharing PoCs that have dual-use export controls.
- May involve government infrastructure or critical services.
Each jurisdiction has different defaults. A researcher in country A testing system in country B disclosing to vendor in country C, publishing in country D, may have to satisfy four legal regimes.
Jurisdiction snapshot
United States — CFAA + DMCA
- CFAA (18 USC §1030) — criminalises “unauthorised access”. Authorisation interpretation has narrowed since Van Buren (2021).
- DMCA §1201 — anti-circumvention; some research exceptions added in 2022.
- State laws vary; California, Illinois have additional consumer-data laws.
- Safe-harbour common in bug bounty contracts; only as strong as the contract terms.
United Kingdom — Computer Misuse Act 1990
- CMA §1 unauthorised access; §2 with intent; §3 modifying.
- Penalties up to 14 years.
- No statutory safe-harbour for security research.
- Industry pressure for CMA reform ongoing.
European Union — NIS2 + national implementations
- NIS2 Directive requires coordinated vulnerability disclosure (CVD) policies at member-state level.
- Cyber Resilience Act (CRA) — manufacturers required to handle vulnerabilities.
- National laws vary in researcher protection: Belgium has explicit safe-harbour; Germany historically restrictive.
Germany — §202a/202b/202c StGB
- §202c “Hacker Paragraph” — criminalises possession of “computer programs intended for committing a crime” (Sec. 202a). Interpreted broadly; has chilled tools.
- Research exceptions in case law but unclear in statute.
- Some researchers don’t publish PoCs out of caution.
Japan — Unauthorised Computer Access Law
- Pre-authorisation requirement is strict.
- Research must be pre-authorised by target.
Australia — Criminal Code Act 1995
- §477.1 unauthorised access; §477.3 production / supply / obtain of data with intent.
- Research carve-outs limited.
Brazil — Carolina Dieckmann Act / LGPD
- Computer-crime law (Lei 12.737) plus data-protection (LGPD).
- Researcher protection limited.
Singapore — Computer Misuse Act + PDPA
- Strict; pre-authorisation expected.
- Government-coordinated VDP through CSA.
South Korea, China, Russia, Iran
- Highly restrictive. Independent research highly risky without explicit government coordination.
Bug bounty contractual safe-harbour
Programs publish a “safe-harbour” or “VDP” section. Common protections:
- Will not initiate civil or criminal action against good-faith researchers acting within scope.
- May indemnify researchers acting within scope.
- Specifies what’s in/out of scope behaviourally (no DoS, no social engineering, no data exfil beyond proof).
Strength varies:
- HackerOne / Bugcrowd standard templates are widely adopted.
- Disclose.io provides a model framework.
- Government VDPs (CISA’s Federal VDP) extend coverage.
These contractual protections often supplement but don’t override criminal statutes; in many countries the government can still prosecute regardless of contract.
Independent research without a bug bounty
If you find a bug in software with no public VDP / bug bounty:
Options
- **Email security@
** — many vendors have an unpublished disclosure inbox. Use clear, factual language; don't threaten disclosure. - CERT coordination — US-CERT (now CISA), JPCERT, KrCERT, others coordinate vendor contact and timeline.
- Coordinated disclosure platforms — Bugcrowd / HackerOne offer mediated disclosure for non-program-vendors.
- Public disclosure — only after exhausting vendor contact, typically with timeline (90 days CERT, or Project Zero 90-day standard).
- Government channel — for critical infrastructure, sometimes coordinated via national CERT.
Risks
- No contract → no safe-harbour even if vendor doesn’t sue.
- Vendor may sue / report to law enforcement — has happened (Coalfire / Iowa case 2019, Aaron Swartz 2013, others).
- Cross-border — vendor pursuing you across borders is more painful than civil suit at home.
Best practices
Before testing
- Verify scope in writing.
- For programs: read every word of the VDP.
- For non-program: pre-authorise via email if at all possible.
- Stay strictly in scope.
During testing
- Minimum-impact validation; don’t exfil beyond proof.
- Document timestamps, IP addresses you used.
- Don’t share PoCs prematurely.
During disclosure
- Use vendor’s published channel.
- Provide actionable details for remediation.
- Agree on timeline (90 days standard, negotiable for criticality).
- Coordinate publication.
After patch
- Publish writeup at the agreed time.
- Credit vendor for cooperation.
- Don’t include weaponisable PoCs if doing so amplifies risk to unpatched systems.
Specific situations
Public-sector / critical infrastructure
- ICS / SCADA / medical-device / utility vulnerabilities have additional reporting obligations (CISA, ICS-CERT, FDA for medical, NCSC for UK CNI).
- Government may impose embargoes for national-security reasons.
- Always coordinate; don’t disclose unilaterally.
National security research
- US ITAR / EAR may restrict sharing certain crypto / exploitation PoCs.
- Wassenaar Arrangement → national export controls on “intrusion software” — varies by country.
- Researchers traveling internationally with offensive tools may face customs issues.
AI / LLM red teaming
- Newer; jurisdiction-specific guidance is sparse.
- Anthropic, OpenAI, Google have published bug bounty terms; treat as binding contract.
- For prompt injection in deployed products: same VDP rules as web apps.
Vulnerability brokering
- Selling vulnerabilities to third-party brokers (ZDI, Zerodium, Crowdfense) is legal in some jurisdictions, restricted in others.
- Government end-customer restrictions may apply.
Jurisdiction-mismatch playbook
For a multi-jurisdiction situation:
- Identify all relevant jurisdictions: your residence, target system location, vendor HQ, publication audience.
- Find the most restrictive that applies; assume that standard.
- Get authorisation in writing from vendor in the most restrictive jurisdiction.
- Avoid publication in jurisdictions where your activity is criminal.
- Consult counsel for high-stakes situations.
Practical resources
- Disclose.io — VDP framework adoption.
- CERT/CC Vulnerability Disclosure Guide.
- Hacker One Hacker Code of Conduct.
- Bugcrowd VRT + standardised disclosure agreements.
- EFF security research guides.
- Lawyers specialising in security research (Tor Ekeland, Marcia Hofmann, others).
For pre-action consultation, EFF and Mozilla maintain attorney contact lists.
Where to start
If you’re new to disclosure:
- Stay inside published bug bounty programs for your first year.
- Read every VDP before testing.
- Build relationships with triagers at programs you target frequently.
- Document everything.
- Have a lawyer’s phone number even if you never call it.
Related
- report-writing
- report-writing-step-by-step
- disclosure-and-comms
- program-scope-reading
- ctf-to-bug-bounty-transition