Multilingual jailbreak chain classifiedGPT-4o · CRITICAL
Unicode evasion governance pack updatedPolicy · v2.3
Singapore NUS team joined research networkUniversity Partner
Claude 3.5 memory leakage mitigation releasedClaude · HIGH
APAC injection taxonomy — 318 patterns validatedIntelligence
Gemini few-shot induction confirmed by two reviewersGemini · HIGH
Provider risk scores recalculated across 5 platformsIntelligence
Copilot Unicode filter bypass — detection integratedCopilot · HIGH
IIT Bombay AI Security Lab joined partner networkUniversity Partner
Base64 double-encode PII bypass — enterprise advisory issuedDLP · CRITICAL
Multilingual jailbreak chain classifiedGPT-4o · CRITICAL
Unicode evasion governance pack updatedPolicy · v2.3
Singapore NUS team joined research networkUniversity Partner
Claude 3.5 memory leakage mitigation releasedClaude · HIGH
APAC injection taxonomy — 318 patterns validatedIntelligence
Gemini few-shot induction confirmed by two reviewersGemini · HIGH
Provider risk scores recalculated across 5 platformsIntelligence
Copilot Unicode filter bypass — detection integratedCopilot · HIGH
IIT Bombay AI Security Lab joined partner networkUniversity Partner
Base64 double-encode PII bypass — enterprise advisory issuedDLP · CRITICAL

Enterprise AI Security Intelligence Network

The Global Operating Layer
for AI Governance Intelligence.

Continuously evolving enterprise protection powered by a global network of security researchers, universities, and governance specialists. Not a report. A living ecosystem.

14,209
Threats Classified
847
Attack Patterns
312
Contributors
38
Bounties Active
5
APAC Jurisdictions
99.7%
Detection Coverage

The Intelligence Flywheel

How Research Becomes
Enterprise Protection

Every validated research contribution directly improves enterprise protection. The cycle is continuous, automatic, and community-powered.

01

Research Submission

Security researchers worldwide submit jailbreak discoveries, injection patterns, and governance gaps through the structured contribution portal.

02

Expert Validation

A team of enterprise security specialists and academic partners verify reproducibility, assess severity, and assign a governance impact score.

03

Threat Classification

Validated findings are indexed into the StreetMP threat taxonomy — categorised by attack family, evasion technique, provider surface, and regulatory impact.

04

Detection Update

The StreetMP neural detection engine is retrained with new adversarial patterns, improving runtime interception accuracy across all deployed tenants.

05

Governance Pack Release

Corresponding policy packs and compliance controls are updated and pushed to all enterprise subscribers — no manual intervention required.

06

Runtime Enforcement

Updated detections activate across every StreetMP enterprise deployment simultaneously — protecting conversations, APIs, and documents at the edge.

07

Enterprise Protected

Every organisation on the StreetMP network benefits from the collective research intelligence — continuously and automatically.

Continuous. Automatic. Community-Powered.

The StreetMP intelligence cycle runs 24/7 — every new discovery makes enterprise AI deployments safer for every organisation on the network.

Live Discoveries

Latest Classified Threats

GPT-4oJailbreak
CRITICAL

Multilingual Role-Switch Bypass

Bahasa–English code-switching defeats English-only moderation. System prompt extraction confirmed.

Detection Integrated
2h ago
Claude 3.5DLP Bypass
HIGH

Base64 Double-Encode PII Bypass

Token-level PII scan circumvented. Model decodes and includes raw PII in structured responses.

Mitigation Released
6h ago
Gemini 1.5Governance
HIGH

Few-Shot System Prompt Extraction

Crafted few-shot examples induce pattern-matching that reveals confidential system prompt fragments.

Validated
1d ago
CopilotDLP Bypass
HIGH

Unicode ZWNJ Evasion — Content Filter

Inserting U+200D between tokens prevents string-matching filters from flagging adversarial payloads.

Governance Pack Updated
2d ago
Llama 3Semantic Attack
HIGH

Hypothetical Framing — Multi-Turn Jailbreak

5+ turn semantic chain progressively erodes safety constraints. No single-turn detection possible.

Under Review
3d ago
AnyPII Leakage
MEDIUM

Malaysian NRIC Transliteration Evasion

Digit look-alike Unicode chars defeat regex-based PII detection for Malaysian National ID numbers.

Detection Integrated
4d ago
Why It Matters

AI Threats Evolve
Faster Than Static Rules

The security landscape around enterprise AI is not stable. Providers update their models. Jailbreaks mutate. Governance requirements expand. New attack families emerge every week. Static detection systems become obsolete before they are even deployed.

The Reality of Enterprise AI Threats

34%

of enterprise AI interactions involve an attempted injection

StreetMP Telemetry, Q1 2026

847

unique jailbreak patterns discovered in the past 12 months

StreetMP Research Registry

6

major AI providers updated their APIs — breaking governance controls

Provider Advisory Monitoring

52%

growth in semantic multi-turn attacks — invisible to static rules

StreetMP Threat Taxonomy

How StreetMP Stays Ahead

Continuously evolving protection

StreetMP's detection and governance systems are powered by continuous research intelligence. Every validated finding from the global research community automatically improves enterprise protections — without manual intervention.

Neural detection modelsUpdated weekly
Governance policy packsUpdated on validated finding
Provider risk profilesRecalculated monthly
Enterprise enforcement rulesDeployed automatically
APAC compliance mappingsUpdated on regulation change
False positive correctionsIntegrated continuously

The platform that gets smarter every single day.

Every research submission, every bounty, every university contribution directly improves the intelligence that protects your enterprise AI deployments. This is not a static product. It is a continuously evolving protection infrastructure.

Global Intelligence Network

Research Intelligence From
Every Corner of the Planet

Researchers from leading universities and enterprises across APAC, Europe, and North America continuously contribute intelligence to the global network.

SingaporeNUS · NTUMumbaiIIT Bombay · IIScTokyoKyoto UniversitySydneyUNSW · MonashSan FranciscoEnterprise HubLondonResearch NetworkKuala LumpurMonash · UTM
7 Active
Research Hubs
352+
Total Contributors
6 Jurisdictions
APAC Coverage
Singapore
NUS · NTU
58 contrib.
Mumbai
IIT Bombay · IISc
44 contrib.
Tokyo
Kyoto University
29 contrib.
Sydney
UNSW · Monash
31 contrib.
San Francisco
Enterprise Hub
52 contrib.
London
Research Network
38 contrib.
Kuala Lumpur
Monash · UTM
19 contrib.

Researcher Identity

Elite Contributor Ecosystem

Researcher reputation is built on impact, not popularity. Contribution quality, validation success, and enterprise adoption drive identity — not social metrics.

Contributor Roles

Verified Researcher

Independent security researchers with ≥10 validated contributions and confirmed identity.

University Researcher

Academic researchers from StreetMP partner institutions.

Governance Specialist

Experts focused on compliance frameworks, policy development, and regulatory mapping.

Enterprise Contributor

Security teams from enterprise organisations with firsthand deployment experience.

Threat Intelligence Analyst

Specialists in attack taxonomy, threat classification, and adversarial intelligence.

Red Team Researcher

Dedicated offensive researchers providing high-quality adversarial test corpora.

What Drives Reputation

Validation Success Rate
Mitigation Influence
Enterprise Adoption
Provider Coverage

Top Contributors — Q1 2026

1
NUS_SecLab
Natl. Univ. Singapore
University Researcher
47
Contributions
96%
Validation
14
Ent. Citations
GPT-4oGeminiClaude
2
r_vasquez
Independent
Threat Intelligence Analyst
38
Contributions
92%
Validation
11
Ent. Citations
ClaudeGPT-4o
3
iitb_ai
IIT Bombay
University Researcher
31
Contributions
90%
Validation
8
Ent. Citations
GeminiCopilotGPT-4o
4
kyoto_sec
Kyoto University
Red Team Researcher
24
Contributions
88%
Validation
5
Ent. Citations
Llama 3Mistral

Research Library

Original Intelligence Reports

THREAT INTELLIGENCEQ1 2026

APAC AI Threat Landscape 2026

14,209 intercepted interactions. Top patterns: prompt injection 38%, credential leakage 29%, PHI exposure 19%.

12 min 24p
Read report
SECURITY RESEARCHMar 2026

Multilingual Jailbreak Taxonomy v3

847 unique jailbreak patterns across 12 languages. Introduces the Semantic Evasion Complexity (SEC) score.

18 min 36p
Read report
ARCHITECTUREFeb 2026

Sovereign AI Runtime Governance

Zero-trust AI governance spec: edge interception, semantic inspection, and cryptographic audit chains.

25 min 48p
Read report
COMPLIANCEDec 2025

AI Governance for APAC Financial Services

Compliance guide for MAS TRM, India PDPB, APPI (Japan) under data sovereignty requirements.

20 min 40p
Read report

University Partners

NUS
IIT Bombay
Kyoto Univ.
UNSW
NTU Singapore
IISc Bangalore
View all programs