EXHIBIT 01OPENING0:00
11th INTERPOL Digital Forensic Expert Group Meeting · 1st International Investigators' Summit
Sovereign National
Forensic AI Agents
United Against Crime — A Global AI Forensic Alliance
Owned nationally · Coordinated through INTERPOL · The human always in command
National Forensic Sciences University · Gandhinagar, India · 13–14 July 2026
Major Engineer Ali Althowaikh — Forensic Science & AI Expert · Ministry of Interior, Kuwait
الاجتماع الحادي عشر لمجموعة خبراء الأدلة الجنائية الرقمية بالإنتربول · القمة الدولية الأولى للمحققين
وكلاء الذكاء الاصطناعي
الجنائي الوطني السيادي
متحدون ضد الجريمة — تحالف جنائي عالمي مدعوم بالذكاء الاصطناعي
ملكية وطنية · تنسيق عبر الإنتربول · القرار دائماً بيد الإنسان
جامعة العلوم الجنائية الوطنية · غاندي ناغار، الهند · ١٣–١٤ يوليو ٢٠٢٦
المقدّم علي الذويخ — خبير علوم جنائية وذكاء اصطناعي · وزارة الداخلية، الكويت
EXHIBIT 02THE CRISIS2:00

The Crisis

Crime is automated, global, and AI-powered

Reactive by design

Evidence goes cold before we examine it. Cyber attackers are in and out in hours. NPS chemists ship product before the analog reaches a scheduling committee.

Threat outpaces the tool

AI malware rewrites its own signature. New drug analogs appear faster than governments schedule them. UNODC: 1,040+ novel substances monitored across 120+ countries by 2023.

Fragmented intelligence

196 nations in silos against criminal networks sharing tools, zero-days, and infrastructure globally. The asymmetry is structural — and it is growing.

Verified scale

FBI IC3 2023: $12.5 billion in documented internet crime losses in the United States alone. $3.94 billion from AI-assisted investment fraud. These are reported figures — actual losses are a multiple.

VERIFIED CASE

Arup Engineering — Hong Kong, February 2024

HK$200M (~USD $25M) transferred after staff attended a deepfake video conference with fabricated executives. Source: Hong Kong Police, Reuters, BBC

حالة مُتحقَّق منها

Arup هونغ كونغ، فبراير ٢٠٢٤

تحويل ٢٠٠ مليون دولار هونغ كونغي (~٢٥ مليون دولار) بعد مؤتمر فيديو مزيّف. المصدر: شرطة هونغ كونغ، Reuters، BBC

VERIFIED CASE

WannaCry — 150 Countries, May 2017

200,000+ systems, NHS hospitals diverted ambulances, $4–8B in damages, attributed to North Korea by US DOJ and UK NCSC

حالة مُتحقَّق منها

WannaCry — ١٥٠ دولة، مايو ٢٠١٧

٢٠٠٠٠٠+ جهاز، مستشفيات NHS حوّلت سيارات الإسعاف، أضرار ٤–٨ مليار دولار. المصدر: DOJ، UK NCSC

الأزمة

الجريمة مُؤتمتة وعالمية ومدعومة بالذكاء الاصطناعي

استجابة لا استباق

الدليل يبرد قبل أن نفحصه. المهاجمون الإلكترونيون يدخلون ويخرجون في ساعات. وكيميائيو المخدّرات يشحنون قبل وصول المركّب إلى لجان الجدولة.

التهديد يسبق الأداة

البرمجيات الخبيثة تُعيد كتابة بصمتها. المركّبات الجديدة تظهر أسرع من جدولتها. UNODC: أكثر من ١٠٤٠ مادة جديدة تحت الرصد في ١٢٠+ دولة بحلول ٢٠٢٣.

استخبارات مجزّأة

١٩٦ دولة في صوامع في مواجهة شبكات إجرامية تتشارك الأدوات وثغرات يوم الصفر والبنية التحتية عالمياً. عدم التماثل هيكلي ويتفاقم.

حجم موثَّق

FBI IC3 2023: ١٢.٥ مليار دولار خسائر جرائم إنترنت موثَّقة في الولايات المتحدة وحدها. ٣.٩٤ مليار من الاحتيال الاستثماري بمساعدة الذكاء الاصطناعي. هذه أرقام مُبلَّغ عنها — الأرقام الفعلية مضاعفة.

VERIFIED CASE

Arup Engineering — Hong Kong, February 2024

HK$200M (~USD $25M) transferred after staff attended a deepfake video conference with fabricated executives. Source: Hong Kong Police, Reuters, BBC

حالة مُتحقَّق منها

Arup هونغ كونغ، فبراير ٢٠٢٤

تحويل ٢٠٠ مليون دولار هونغ كونغي (~٢٥ مليون دولار) بعد مؤتمر فيديو مزيّف. المصدر: شرطة هونغ كونغ، Reuters، BBC

VERIFIED CASE

WannaCry — 150 Countries, May 2017

200,000+ systems, NHS hospitals diverted ambulances, $4–8B in damages, attributed to North Korea by US DOJ and UK NCSC

حالة مُتحقَّق منها

WannaCry — ١٥٠ دولة، مايو ٢٠١٧

٢٠٠٠٠٠+ جهاز، مستشفيات NHS حوّلت سيارات الإسعاف، أضرار ٤–٨ مليار دولار. المصدر: DOJ، UK NCSC

EXHIBIT 03CRIMINAL AI4:30

How Criminals Use AI

Six active vectors — every one documented from primary sources
🎭

Synthetic Identity & Fraud

LLMs generate flawless spear-phishing in 50+ languages. Deepfakes impersonate executives. Voice clones bypass call-centre authentication. ⚑ Arup $25M deepfake video call (HK Police, Feb 2024) ⚑ Ferrari CEO voice-clone attempt (Bloomberg, 2024)

💻

AI-Powered Cyber Attacks

Polymorphic malware rewrites its signature per target. AI finds zero-days faster than patches deploy. Ransomware-as-a-Service with automated targeting. ⚑ WormGPT — dark web LLM, no guardrails (SlashNext, Jul 2023) ⚑ LockBit 3.0 RaaS — $120M+ extorted (FBI/NCA, 2023–24)

💊

Novel Synthetic Drugs (NPS)

Generative chemistry designs analogs one molecule beyond scheduling — legal until governments react. Fentanyl analogs, designer cannabinoids. ⚑ 1,040+ NPS in 120+ countries (UNODC 2023) ⚑ 50+ new substances in Europe in 2023 alone (EMCDDA 2023)

🌐

Dark Web Automation

AI marketplaces auto-match buyers and suppliers. Encrypted comms auto-translate across language barriers. Logistics optimised to evade postal interdiction. ⚑ 288 vendors, $53.4M — Operation SpecTor (Europol/DOJ, May 2023) ⚑ 34 LockBit servers seized — Operation Cronos (NCA/FBI, Feb 2024)

💸

Financial Crime at Scale

AI traces clean crypto-mixing routes. Pig-butchering scams run by LLM chatbots. Shell-company networks auto-generated. ⚑ $24.2B illicit crypto in 2023 (Chainalysis 2024) ⚑ $3.94B investment fraud losses (FBI IC3 2023)

📣

Disinformation & Influence

Synthetic media floods platforms before fact-checkers respond. State actors confirmed using ChatGPT for influence operations. ⚑ Slovakia election deepfake audio, 48h before polls (AFP, Sep 2023) ⚑ Russia/China/Iran/N. Korea accounts terminated (OpenAI, May 2024)

كيف يستخدم المجرمون الذكاء الاصطناعي

ستة محاور نشطة — كل واحد موثَّق من مصادر أولية
🎭

الهوية المُصنَّعة والاحتيال

نماذج لغوية تُولِّد رسائل تصيّد مثالية. تزييف عميق ينتحل شخصية المديرين. استنساخ صوت يتجاوز مصادقة مراكز الاتصال. ⚑ Arup ٢٥M$ (شرطة هونغ كونغ، فبراير ٢٠٢٤) ⚑ محاولة استنساخ صوت مدير Ferrari (Bloomberg، ٢٠٢٤)

💻

الهجمات الإلكترونية بالذكاء الاصطناعي

برمجيات تُعيد كتابة بصمتها. ذكاء اصطناعي يجد الثغرات أسرع من التحديثات. برامج فدية كخدمة. ⚑ WormGPT (SlashNext، يوليو ٢٠٢٣) ⚑ LockBit 3.0 — ١٢٠M$+ (FBI/NCA، ٢٠٢٣–٢٤)

💊

المخدّرات المُصنَّعة الجديدة

أدوات الكيمياء التوليدية تُصمِّم مركّبات جديدة بجزيء واحد خارج الجدولة. ⚑ أكثر من ١٠٤٠ NPS (UNODC ٢٠٢٣) ⚑ أكثر من ٥٠ مادة جديدة في أوروبا عام ٢٠٢٣ (EMCDDA)

🌐

أتمتة الشبكة المظلمة

أسواق مدعومة بالذكاء الاصطناعي. اتصالات مُشفَّرة تُترجم تلقائياً. ⚑ ٢٨٨ بائعاً، ٥٣.٤M$ — عملية SpecTor (Europol، مايو ٢٠٢٣) ⚑ ٣٤ خادماً — عملية Cronos (NCA/FBI، فبراير ٢٠٢٤)

💸

الجريمة المالية بالجملة

ذكاء اصطناعي يتتبّع مسارات التعمية. عمليات نصب عاطفي بروبوتات محادثة. ⚑ ٢٤.٢ مليار عملات مشفّرة غير مشروعة (Chainalysis ٢٠٢٤) ⚑ ٣.٩٤ مليار احتيال استثماري (FBI IC3 ٢٠٢٣)

📣

التضليل والتأثير

وسائط مُصنَّعة تُغرق المنصات. دول مؤكَّدة استخدمت ChatGPT. ⚑ تزييف صوتي في انتخابات سلوفاكيا (AFP، سبتمبر ٢٠٢٣) ⚑ روسيا والصين وإيران وكوريا الشمالية (OpenAI، مايو ٢٠٢٤)

EXHIBIT 04CYBER THREATS7:00

AI-Powered Cyber Attacks

Every example below is documented from primary sources

Polymorphic & evasive malware

WormGPT (SlashNext, July 2023) — a purpose-built dark-web LLM with no ethical guardrails, sold for malware generation and BEC attacks. AI-generated malware mutates its signature on every execution. Traditional signature-based AV fails completely.

Automated vulnerability discovery

AI scans millions of code lines and network configs faster than any human red team. Critical infrastructure is the documented primary target. Patch cycles cannot keep pace with AI-assisted discovery.

Spear-phishing at industrial scale

LLMs remove the one traditional tell: language errors. Microsoft Digital Defense Report 2023: 4,000+ password attacks per second against Microsoft cloud services. Volume is now machine-scale.

Ransomware-as-a-Service + AI

LockBit 3.0 used AI-optimised target selection. $120M+ extorted from 2,000+ victims before Operation Cronos (NCA/FBI/Europol, Feb 2024) seized 34 servers and froze 200+ crypto accounts.

VERIFIED CASE

Colonial Pipeline — May 2021

DarkSide RaaS. 6-day shutdown of the largest US East Coast fuel pipeline. $4.4M ransom paid. FBI recovered $2.3M. Source: US DOJ, FBI

حالة مُتحقَّق منها

Colonial Pipeline، مايو ٢٠٢١

DarkSide RaaS. إغلاق ٦ أيام. فدية ٤.٤ مليون دولار. المصدر: DOJ، FBI

VERIFIED CASE

Operation Cronos — February 2024

NCA/FBI/Europol joint operation. 34 LockBit servers seized across 10+ countries. 200+ crypto accounts frozen. Source: NCA, Europol, FBI joint announcement

حالة مُتحقَّق منها

عملية Cronos، فبراير ٢٠٢٤

NCA/FBI/Europol. ٣٤ خادماً. ٢٠٠+ حساب مُجمَّد. المصدر: NCA، Europol، FBI

الهجمات الإلكترونية المدعومة بالذكاء الاصطناعي

كل مثال أدناه موثَّق من مصادر أولية

برمجيات خبيثة متعدّدة الأشكال

WormGPT (SlashNext، يوليو ٢٠٢٣) — نموذج لغوي للشبكة المظلمة بلا ضمانات أخلاقية، يُباع لصنع البرمجيات الخبيثة. البرمجيات تُغيِّر بصمتها في كل تنفيذ. مكافحة الفيروسات التقليدية تفشل تماماً.

اكتشاف ثغرات تلقائي

الذكاء الاصطناعي يفحص ملايين أسطر الكود أسرع من أي فريق اختبار اختراق بشري. البنية التحتية الحرجة هي الهدف الأول الموثَّق. دورات التحديث لا تُجاري سرعة الاكتشاف بالذكاء الاصطناعي.

تصيّد موجَّه بالجملة

النماذج اللغوية تُزيل الدليل التقليدي الوحيد: الأخطاء اللغوية. Microsoft DDIR 2023: أكثر من ٤٠٠٠ هجوم كلمة مرور في الثانية. الحجم الآن على مقياس الآلة.

برامج الفدية كخدمة + الذكاء الاصطناعي

LockBit 3.0 استخدم استهدافاً مُحسَّناً. ١٢٠M$+ من ٢٠٠٠+ ضحية قبل عملية Cronos (NCA/FBI/Europol، فبراير ٢٠٢٤) التي أسقطت ٣٤ خادماً.

VERIFIED CASE

Colonial Pipeline — May 2021

DarkSide RaaS. 6-day shutdown of the largest US East Coast fuel pipeline. $4.4M ransom paid. FBI recovered $2.3M. Source: US DOJ, FBI

حالة مُتحقَّق منها

Colonial Pipeline، مايو ٢٠٢١

DarkSide RaaS. إغلاق ٦ أيام. فدية ٤.٤ مليون دولار. المصدر: DOJ، FBI

VERIFIED CASE

Operation Cronos — February 2024

NCA/FBI/Europol joint operation. 34 LockBit servers seized across 10+ countries. 200+ crypto accounts frozen. Source: NCA, Europol, FBI joint announcement

حالة مُتحقَّق منها

عملية Cronos، فبراير ٢٠٢٤

NCA/FBI/Europol. ٣٤ خادماً. ٢٠٠+ حساب مُجمَّد. المصدر: NCA، Europol، FBI

EXHIBIT 05THE SHIFT9:00

The Shift

From isolated forensics to sovereign forensic intelligence
NOT from 'what happened?' to 'who will offend next?' — that road leads to bias, wrongful targeting, inadmissible evidence, and legal liability that collapses prosecutions.
THE AGENT

Scans millions of artifacts. Detects anomalies and criminal patterns. Surfaces a verified lead with a documented confidence score and error rate.

THE INVESTIGATOR

Judges the lead. Corroborates by accepted method. Bears professional responsibility for the conclusion. Testifies under oath.

The machine narrows the haystack. The human renders the judgment. The court decides.

التحوّل

من أدلة جنائية معزولة إلى ذكاء جنائي سيادي مترابط
ليس من «ماذا حدث؟» إلى «من سيرتكب الجريمة؟» — فهذا طريق التحيّز والاستهداف الظالم والأدلة غير المقبولة والمسؤولية القانونية التي تُسقط الملاحقات.
الوكيل الذكي

يفحص ملايين الآثار. يكشف الأنماط الإجرامية. يرفع دليلاً مبدئياً مع درجة ثقة موثَّقة ومعدل خطأ منشور.

المحقّق الإنساني

يحكم على الدليل المبدئي. يتحقّق بالطريقة المعتمدة. يتحمّل المسؤولية المهنية عن الاستنتاج. يشهد أمام القضاء.

الآلة تضيّق دائرة البحث. الإنسان يُصدر الحكم. المحكمة تقرّر.
EXHIBIT 06THE RACE11:00

We Are in a Race

Every criminal capability — a documented counter · Every counter — a named standard

AI used by the criminal

ذكاء يستخدمه المجرم

  • Deepfakes, voice clones, synthetic identities (Arup $25M, 2024)
    تزييف عميق واستنساخ صوت (Arup ٢٥M$، ٢٠٢٤)
  • Polymorphic malware & zero-day AI (WormGPT, Jul 2023)
    برمجيات متعدّدة الأشكال (WormGPT، يوليو ٢٠٢٣)
  • Novel drug analog design — 1,040+ NPS (UNODC 2023)
    تصميم مخدّرات جديدة — ١٠٤٠+ NPS (UNODC ٢٠٢٣)
  • Ransomware-as-a-Service (LockBit: $120M+ extorted)
    برامج فدية كخدمة (LockBit: ١٢٠M$+)
  • LLM-powered fraud & pig-butchering ($3.94B, FBI 2023)
    احتيال بنماذج لغوية (٣.٩٤ مليار، FBI ٢٠٢٣)
  • AI disinformation — state-level (OpenAI, May 2024)
    تضليل بمستوى الدول (OpenAI، مايو ٢٠٢٤)
VS

AI used by us — the forensic agent

ذكاء نستخدمه نحن — الوكيل الجنائي

  • Synthetic media detection & biometric verification (NIST FRVT validated)
    كشف الوسائط المُصنَّعة والتحقّق الحيوي (NIST FRVT)
  • Behavioural malware analysis — behaviour, not signatures
    تحليل سلوك البرمجيات — السلوك لا البصمات
  • Spectral fingerprinting of drug analogs (UNODC methodology)
    بصمة طيفية لمماثلات المخدّرات (منهجية UNODC)
  • Blockchain & crypto graph tracing (Chainalysis methodology)
    تتبّع البلوكشين والعملات المشفّرة (منهجية Chainalysis)
  • Transaction pattern analysis & fraud attribution (FBI IC3 framework)
    تحليل أنماط المعاملات (إطار FBI IC3)
  • Coordinated criminal-network behaviour detection (INTERPOL–UNICRI scope)
    كشف سلوك الشبكات الإجرامية المنسَّق (نطاق INTERPOL–UNICRI)
The criminal operates globally as one network. United — sovereign in each nation, coordinated through INTERPOL — we match that unity without surrendering sovereignty.
المجرم يعمل عالمياً كشبكة واحدة. متّحدين — بسيادة كل دولة وتنسيق عبر الإنتربول — نُجاري هذه الوحدة دون التنازل عن السيادة.
EXHIBIT 07SOVEREIGNTY12:30

Sovereign by Design

National ownership · Infrastructure-tiered access · Global coordination

National ownership — no exceptions

Models trained on national data, processed locally, offline-capable. Evidence never crosses a border unless the nation authorises it explicitly. If the coordination network is cloud-dependent, it is itself an attack surface. Ours is not.

Infrastructure-tiered access — not equal access

We do not promise identical capability to every nation immediately. Nations participate at their infrastructure readiness level. A Tier 1 nation runs full capability. A Tier 3 nation starts with spectral drug analysis and grows. Honest about this — or we lose trust.

Coordination, not control — Constitution Article 2

INTERPOL is a coordination body — not an enforcement authority. The INTERPOL Constitution, Article 2, makes this explicit and binding. Enforcement stays national. Always. No exceptions.

Governance framework — four named standards

INTERPOL–UNICRI Responsible AI for Law Enforcement (2023) · NIST FRVT · ISO/IEC 17025 · Daubert / civil law equivalents per jurisdiction.

السيادة بالتصميم

ملكية وطنية · وصول متدرّج حسب البنية التحتية · تنسيق عالمي

الملكية الوطنية — بلا استثناءات

نماذج تُدرَّب على البيانات الوطنية وتُعالَج محلياً وتعمل دون اتصال. الدليل لا يعبر الحدود إلا بتفويض وطني صريح. إذا كانت شبكة التنسيق تعتمد على السحابة، فهي بحد ذاتها سطح هجوم. شبكتنا ليست كذلك.

وصول متدرّج — لا وصول متكافئ فوري

لا نعد بقدرات متطابقة فورية لكل دولة. الدول تشارك وفق مستوى جاهزيتها التقنية. الفئة الأولى تُشغِّل القدرة الكاملة. الفئة الثالثة تبدأ بتحليل المخدّرات الطيفي وتنمو. الصدق في هذا — أو نفقد الثقة.

تنسيق لا تحكّم — المادة ٢ من الميثاق

الإنتربول جهة تنسيق — لا سلطة إنفاذ. ميثاق الإنتربول، المادة ٢، ينصّ على ذلك صراحةً وبشكل ملزم. الإنفاذ يبقى وطنياً. دائماً. بلا استثناءات.

إطار الحوكمة — أربعة معايير مُسمَّاة

INTERPOL–UNICRI للذكاء الاصطناعي المسؤول (٢٠٢٣) · NIST FRVT · ISO/IEC 17025 · Daubert وما يعادله في القانون المدني لكل ولاية قضائية.

EXHIBIT 08FORENSIC CAPABILITIES14:30

Seven Forensic Functions

NIST FRVT benchmark conditions · Documented error rates per demographic group · Every output is a lead
01

Deepfake & synthetic media

Manipulated video, audio, and synthetic identity detection

02

Image & facial comparison

Biometric matching — validated under NIST FRVT · Demographic error rates published

03

Audio & voice verification

Multilingual transcription · Speaker verification with documented false acceptance rate

04

CCTV / NVR analysis

Real-time pattern review · Anomaly detection with human review gate

05

Signature verification

Handwriting and document authentication · ISO 17025 chain of custody

06

ID & passport integrity

Forgery detection across document classes · Known error rate per document type

07

Novel substance recognition

Spectral fingerprinting of NPS analogs · UNODC methodology · Cannot schedule — can fingerprint

Benchmark ≠ field. Demographic bias documented per NIST FRVT 2019. Field error rates published before any deployment. A lead is not evidence. Source: NIST FRVT — pages.nist.gov/frvt

سبع وظائف جنائية

ظروف اختبار NIST FRVT · معدلات خطأ موثَّقة لكل مجموعة ديموغرافية · كل ناتج دليل مبدئي
٠١

كشف التزييف العميق

التعرّف على الفيديو والصوت المُلاعَب بهما والهوية المُصنَّعة

٠٢

مقارنة الصور والوجوه

مطابقة حيوية مُتحقَّق منها · NIST FRVT · معدلات خطأ ديموغرافية منشورة

٠٣

التحقّق من الصوت

تفريغ متعدّد اللغات · التحقّق من المتحدّث مع معدل قبول خاطئ موثَّق

٠٤

تحليل كاميرات المراقبة

مراجعة الأنماط الآنية · كشف الشذوذات مع بوّابة مراجعة بشرية

٠٥

مطابقة التواقيع

توثيق خط اليد والوثائق · سلسلة حيازة ISO 17025

٠٦

سلامة الهويات والجوازات

كشف التزوير · معدل خطأ معروف لكل نوع وثيقة

٠٧

التعرّف على المواد الجديدة

بصمة طيفية لمماثلات NPS · منهجية UNODC · لا نجدول — نُبصّم

الاختبار ≠ الميدان. التحيّز الديموغرافي موثَّق وفق NIST FRVT 2019. معدلات الخطأ الميدانية تُنشر قبل أي نشر. الدليل المبدئي ليس دليلاً قاطعاً. المصدر: pages.nist.gov/frvt
EXHIBIT 09CYBER CAPABILITIES16:00

Cyber & Digital Intelligence

Our counter-arsenal · Behaviour, not signatures · Human-reviewed at every gate

Behavioural malware analysis

We analyse behaviour — not signatures. Signatures fail against polymorphic threats. Behaviour does not mutate. Every sample produces an investigative lead with a documented confidence score.

Network anomaly & intrusion detection

Models trained on national traffic baselines flag deviations in real time. Lateral movement, exfiltration, and C2 channels surface as leads. No automated action — human review required before any enforcement step.

Cryptocurrency & blockchain tracing

Graph-analysis traces funds through mixing and chain-hopping. Chainalysis methodology — the same approach supporting international prosecutions. Surfaces the human wallet at the chain end.

Dark web intelligence

Automated monitoring of darknet markets and criminal forums — tracking threat actors, NPS analogs, leaked credentials, and planned attacks — within INTERPOL–UNICRI defined scope. Not political monitoring.

VERIFIED OPERATION

INTERPOL Operation First Light 2024 — July 2024

61 countries · 3,950 arrested · 14,643 accounts frozen · $257M seized · AI-assisted fraud targeted. Source: INTERPOL press release, July 2024

عملية مُتحقَّق منها

عملية INTERPOL First Light 2024 — يوليو ٢٠٢٤

٦١ دولة · ٣٩٥٠ موقوفاً · ٢٥٧ مليون دولار مُضبوطة. المصدر: الإنتربول، يوليو ٢٠٢٤

VERIFIED OPERATION

Operation SpecTor — Europol / DOJ, May 2023

9 countries · 288 vendors arrested · 850kg drugs seized · $53.4M seized. Source: Europol, US DOJ

عملية مُتحقَّق منها

عملية SpecTor — Europol / DOJ، مايو ٢٠٢٣

٩ دول · ٢٨٨ بائعاً · ٥٣.٤ مليون دولار. المصدر: Europol، DOJ الأمريكي

الاستخبارات الإلكترونية والرقمية

ترسانتنا المضادة · السلوك لا البصمات · مراجعة بشرية عند كل بوّابة

تحليل سلوك البرمجيات الخبيثة

نحلّل السلوك لا البصمات. البصمات تفشل أمام التهديدات المتحوِّلة. السلوك لا يتحوَّل. كل عيّنة تُنتج دليلاً تحقيقياً مع درجة ثقة موثَّقة.

كشف شذوذات الشبكة

نماذج مُدرَّبة على خطوط أساسية وطنية ترصد الانحرافات آنياً. الحركة الجانبية والتسريب وقنوات C2 تظهر كدلائل. لا إجراء تلقائي — مراجعة بشرية مطلوبة قبل أي خطوة إنفاذ.

تتبّع العملات المشفّرة

تحليل الرسم البياني يتتبّع الأموال. منهجية Chainalysis — نفس الأسلوب الداعم للملاحقات الدولية. تكشف المحفظة البشرية في نهاية السلسلة.

استخبارات الشبكة المظلمة

مراقبة مُؤتمتة لأسواق الدارك ويب — ضمن النطاق المحدَّد بـ INTERPOL–UNICRI. ليست مراقبة سياسية.

VERIFIED OPERATION

INTERPOL Operation First Light 2024 — July 2024

61 countries · 3,950 arrested · 14,643 accounts frozen · $257M seized · AI-assisted fraud targeted. Source: INTERPOL press release, July 2024

عملية مُتحقَّق منها

عملية INTERPOL First Light 2024 — يوليو ٢٠٢٤

٦١ دولة · ٣٩٥٠ موقوفاً · ٢٥٧ مليون دولار مُضبوطة. المصدر: الإنتربول، يوليو ٢٠٢٤

VERIFIED OPERATION

Operation SpecTor — Europol / DOJ, May 2023

9 countries · 288 vendors arrested · 850kg drugs seized · $53.4M seized. Source: Europol, US DOJ

عملية مُتحقَّق منها

عملية SpecTor — Europol / DOJ، مايو ٢٠٢٣

٩ دول · ٢٨٨ بائعاً · ٥٣.٤ مليون دولار. المصدر: Europol، DOJ الأمريكي

EXHIBIT 10THE PRINCIPLE17:30
A lead is not evidence.
INTERPOL–UNICRI Responsible AI for Law Enforcement (2023) establishes this as the binding governance standard. The agent surfaces the lead with a confidence score and a documented error rate. The investigator corroborates by accepted method and bears professional responsibility for the conclusion. The court decides admissibility. The machine never takes the stand.
INTERPOL–UNICRI RESPONSIBLE AI FOR LAW ENFORCEMENT · 2023
الدليل المبدئي ليس دليلاً قاطعاً.
إطار INTERPOL–UNICRI للذكاء الاصطناعي المسؤول في إنفاذ القانون (٢٠٢٣) يُرسّخ هذا كمعيار حوكمة ملزم. الوكيل يرفع الدليل المبدئي مع درجة ثقة ومعدل خطأ موثَّق. المحقّق يتحقّق بالطريقة المعتمدة ويتحمّل المسؤولية المهنية عن الاستنتاج. المحكمة تقرّر قبول الأدلة. الآلة لا تعتلي منصّة الشهود.
INTERPOL–UNICRI الذكاء الاصطناعي المسؤول · ٢٠٢٣
EXHIBIT 11PROOF OF METHOD18:30

The Learning Loop — Tested

The methodology that matters: build · measure · document limits · refine
NIST FRVT
the validation gold standard
ISO 17025
the chain-of-custody standard

The learning loop is not theoretical. I built and tested it on one of the hardest prediction problems available — football outcomes — to prove the methodology: data in, model trains, accuracy measured, limits documented, model refined. The discipline proved. The application is forensics.

Forensic AI recognises what an artifact already IS. Not prediction — recognition. That is a tractable signal. NIST FRVT validates face recognition systems to this standard. We align to NIST FRVT — not to football results. The benchmark that matters in this room is the one a court will accept.

The discipline this field demands: every model publishes its error rate before it enters a courtroom. Every. Single. Time. A method you cannot characterise is a method you cannot defend. We characterise ours first.

حلقة التعلّم — مُختبَرة

المنهجية التي تهمّ: ابنِ · قِس · وثِّق الحدود · حسِّن
NIST FRVT
المعيار الذهبي للتحقّق
ISO 17025
معيار سلسلة الحيازة

حلقة التعلّم ليست نظرية. بنيتُها واختبرتُها على أصعب مسائل التنبّؤ — نتائج كرة القدم — لإثبات المنهجية: بيانات، تدريب، قياس، توثيق، تحسين. المنهجية أُثبتت. التطبيق هو الجنائيات.

الذكاء الاصطناعي الجنائي يتعرّف على ما هو كائن فعلاً. ليس تنبّؤاً — تعرّف. إشارة قابلة للتحقّق. NIST FRVT هو معيار التحقّق الذي تقبله المحاكم. نلتزم به.

الانضباط الذي تتطلّبه هذه المهنة: كل نموذج ينشر معدل خطئه قبل قاعة المحكمة. كل مرة. منهج لا تستطيع توصيفه لا تستطيع الدفاع عنه. نُوصِّف نظامنا أولاً.

EXHIBIT 12GOVERNANCE · PART 120:30

Human in Command

Four guardrails · Four named standards · Zero exceptions
01

Human-in-command at every step — INTERPOL–UNICRI 2023

The agent never arrests, charges, attributes a cyber attack, or decides. It surfaces a lead. The investigator corroborates and bears professional responsibility. The court decides admissibility. Three tiers of accountability — no tier skipped.

02

Demographic bias testing — NIST FRVT 2019 mandatory

NIST FRVT 2019 documented significantly higher error rates for some demographic groups. We publish error rates per demographic group before any deployment. If a system fails the demographic threshold — it does not deploy. No exceptions.

03

Legal liability framework — ISO/IEC 17025

Developer documents error rates per ISO 17025. Investigator who acts on the lead bears professional corroboration responsibility. No automated enforcement action is permitted. Evidence collapses without documented chain of custody.

04

Cross-jurisdictional admissibility — jurisdiction-by-jurisdiction review

Output is a lead, not evidence. Admissibility is tested at corroboration stage in each jurisdiction. We align to Daubert (US/common law), equivalent civil law standards (France/Germany/GCC), and local legal frameworks — reviewed before deployment in each nation.

القرار بيد الإنسان

أربع ضمانات · أربعة معايير مُسمَّاة · لا استثناءات
01

الإنسان في القيادة عند كل خطوة — INTERPOL–UNICRI ٢٠٢٣

الوكيل لا يعتقل ولا يتّهم ولا يُسنِد هجوماً ولا يقرّر. يرفع دليلاً مبدئياً. المحقّق يتحقّق ويتحمّل المسؤولية المهنية. المحكمة تقرّر قبول الأدلة. ثلاثة مستويات مساءلة — لا مستوى يُتخطَّى.

02

اختبار التحيّز الديموغرافي — NIST FRVT 2019 إلزامي

NIST FRVT 2019 وثَّق معدلات خطأ أعلى بكثير لبعض المجموعات الديموغرافية. ننشر معدلات الخطأ لكل مجموعة قبل أي نشر. إذا فشل النظام في العتبة الديموغرافية — لا يُنشر. لا استثناءات.

03

إطار المسؤولية القانونية — ISO/IEC 17025

المطوِّر يُوثِّق معدلات الخطأ وفق ISO 17025. المحقق الذي يتصرّف بناءً على الدليل المبدئي يتحمّل المسؤولية المهنية عن التحقّق. لا إجراء إنفاذ تلقائي مسموح. الدليل ينهار دون سلسلة حيازة موثَّقة.

04

قبول الأدلة عبر الولايات القضائية — مراجعة قانون قضاء بقضاء

الناتج دليل مبدئي لا دليل قاطع. القبول يُختبر عند مرحلة التحقّق في كل ولاية قضائية. نلتزم بـ Daubert (القانون الأمريكي)، والمعادلات في القانون المدني (فرنسا، ألمانيا، دول الخليج)، والأطر القانونية المحلية — تُراجَع قبل النشر في كل دولة.

EXHIBIT 13GOVERNANCE · PART 222:00

Known Risks — Our Position

We name our vulnerabilities before you do · Four active mitigations
01

Adversarial AI — criminals will attack the detector itself

An adversary who knows we use facial recognition will craft adversarial examples — pixel-level manipulations that fool the model while appearing normal to the human eye. Mitigation: we conduct red-team exercises against our own detectors using NIST adversarial testing protocols. We test the attack surface of the system, not just its accuracy on clean data.

02

Data protection — GDPR Article 46 and equivalents

Cross-border data movement requires explicit legal basis under GDPR Article 46 and equivalent national privacy laws. Our architecture processes locally by default — no data crosses a border without a specific legal instrument. Data minimisation by design. Right to explanation for any adverse outcome. Privacy Impact Assessment before each national deployment.

03

Definitional safeguards — criminal networks, not political speech

'Coordinated inauthentic behaviour' is a term that can be misused. We define our scope precisely: accounts and networks linked to documented criminal infrastructure, financial fraud, and known criminal operations — per INTERPOL–UNICRI framework. This capability is not a tool for monitoring political opposition, religious groups, or civil society. That boundary is technical, legal, and contractual.

04

False positive cost — a wrongful lead has consequences

A false positive in biometrics means an innocent person under investigation. We require: (a) error rates documented before deployment, (b) no action on a single-source AI lead alone, (c) documented corroboration before any arrest or charge. The cost of false positives is not statistical — it is a human being's liberty.

المخاطر المعروفة — موقفنا

نُسمّي ثغراتنا قبل أن تُسمّيها · أربعة تخفيفات نشطة
01

الذكاء الاصطناعي المعادي — المجرمون سيهاجمون الكاشف نفسه

الخصم الذي يعلم أننا نستخدم التعرّف على الوجه سيصنع أمثلة معادية — تلاعب على مستوى البكسل يخدع النموذج مع بدو طبيعي للعين البشرية. التخفيف: نُجري تمارين الفريق الأحمر ضد كاشفاتنا الخاصة باستخدام بروتوكولات الاختبار المعادي لـ NIST. نختبر سطح الهجوم للنظام، لا دقّته على بيانات نظيفة فحسب.

02

حماية البيانات — المادة ٤٦ من GDPR وما يعادلها

نقل البيانات عبر الحدود يتطلّب أساساً قانونياً صريحاً بموجب المادة ٤٦ من GDPR وقوانين الخصوصية الوطنية المعادلة. بنيتنا تُعالج محلياً افتراضياً — لا بيانات تعبر حدوداً دون صكّ قانوني محدّد. تقليص البيانات بالتصميم. الحق في التفسير لأي نتيجة سلبية. تقييم أثر على الخصوصية قبل كل نشر وطني.

03

ضمانات تعريفية — الشبكات الإجرامية لا الخطاب السياسي

«السلوك غير الأصيل المنسَّق» مصطلح قابل للإساءة. نُحدِّد نطاقنا بدقّة: الحسابات والشبكات المرتبطة بالبنية التحتية الإجرامية الموثَّقة والاحتيال المالي والعمليات الإجرامية المعروفة — وفق إطار INTERPOL–UNICRI. هذه القدرة ليست أداةً لمراقبة المعارضة السياسية أو الجماعات الدينية أو المجتمع المدني. هذا الحدّ تقني وقانوني وتعاقدي.

04

تكلفة الإيجابي الكاذب — الدليل المبدئي الخاطئ له عواقب

الإيجابي الكاذب في القياسات الحيوية يعني شخصاً بريئاً تحت التحقيق. نشترط: (أ) توثيق معدلات الخطأ قبل النشر، (ب) عدم اتخاذ أي إجراء على دليل مبدئي من مصدر الذكاء الاصطناعي وحده، (ج) تحقّق موثَّق قبل أي اعتقال أو اتهام. تكلفة الإيجابيات الكاذبة ليست إحصائية — إنها حرية إنسان.

EXHIBIT 14COORDINATION23:30

Coordination Without Surrender

INTERPOL First Light 2024 — 61 countries, $257M, one coordinated operation — the model that works

Validated intelligence only — never raw data

Only vetted, minimised intelligence moves through INTERPOL channels. Never raw national biometric data, never personal records. The intelligence product, not the source data.

Each nation controls its own border — by design

Data sovereignty is technical, not just policy. Local processing by default. No cross-border transfer without an explicit national authorisation instrument.

Case linkage — the operational proof

Operation First Light 2024 (INTERPOL): 61 countries, simultaneous action, 3,950 arrested, $257M seized. Operation Cronos (NCA/FBI/Europol): 10+ countries, 34 LockBit servers seized in one coordinated strike. Coordination works. We institutionalise what already works.

No trade-off — sovereignty and cooperation are not opposites

Every nation that joined First Light retained full sovereignty. Every nation that joined Cronos acted under its own legal framework. The coordination produced what no single nation could produce alone.

تنسيق دون تنازل

عملية INTERPOL First Light 2024 — ٦١ دولة و٢٥٧ مليون دولار — النموذج الذي يعمل

معلومة مُتحقَّقة فقط — لا بيانات خام

المعلومة المُدقَّقة المُقلَّصة فقط تتحرّك عبر قنوات الإنتربول. لا بيانات حيوية وطنية خام، لا سجلات شخصية. الناتج الاستخباراتي، لا البيانات المصدرية.

كل دولة تتحكّم في حدودها الخاصة — بالتصميم

سيادة البيانات تقنية لا مجرّد سياسة. معالجة محلية افتراضياً. لا نقل عبر حدود دون صكّ تفويض وطني صريح.

ربط القضايا — الإثبات التشغيلي

عملية First Light 2024 (الإنتربول): ٦١ دولة، عمل متزامن، ٣٩٥٠ موقوفاً، ٢٥٧ مليون دولار. عملية Cronos (NCA/FBI/Europol): ١٠+ دول، ٣٤ خادماً في ضربة واحدة. التنسيق يعمل. نحن نُؤسِّس ما يعمل بالفعل.

لا مفاضلة — السيادة والتعاون ليسا نقيضين

كل دولة انضمّت إلى First Light احتفظت بسيادتها الكاملة. كل دولة انضمّت إلى Cronos تصرّفت وفق إطارها القانوني الخاص. التنسيق أنتج ما لا تستطيع دولة واحدة إنتاجه وحدها.

EXHIBIT 15ROADMAP24:30

A Realistic Path

Each phase earned — not promised · Governance peer-reviewed at every step
2026

Validate

3–5 willing nations. Full ISO/IEC 17025 and NIST FRVT validation. Demographic bias testing. Document field error rates. Publish results openly. Earn the right to expand.

2027–28

Expand

20–30 nations with demonstrated infrastructure readiness. Regional training hubs. Peer-reviewed governance published. First live cross-border intelligence exchange under INTERPOL coordination.

2029–30

Federate

Broader INTERPOL coordination channels once governance is independently audited and validated by this expert group and peer institutions.

2031+

Continuous

Annual retraining against emerging criminal AI. Governance reviewed and published annually. Error rates updated publicly. The race never ends — neither does the audit.

مسار واقعي

كل مرحلة تُكسَب — لا تُوعَد · حوكمة مُراجَعة من الأقران عند كل خطوة
٢٠٢٦

التحقّق

٣–٥ دول راغبة. تحقّق كامل وفق ISO/IEC 17025 وNIST FRVT. اختبار التحيّز الديموغرافي. توثيق معدلات الخطأ الميدانية. نشر النتائج علناً. كسب الحق في التوسّع.

٢٠٢٧–٢٠٢٨

التوسّع

٢٠–٣٠ دولة بجاهزية بنية تحتية مُثبَتة. مراكز تدريب إقليمية. حوكمة مُراجَعة من الأقران منشورة. أول تبادل استخباراتي حيّ عبر الحدود تحت تنسيق الإنتربول.

٢٠٢٩–٢٠٣٠

الاتحاد

قنوات تنسيق أوسع عبر الإنتربول بعد مراجعة الحوكمة باستقلالية وتحقّق مجموعة الخبراء هذه والمؤسسات المستقلة.

٢٠٣١+

الاستمرارية

إعادة تدريب سنوية ضد الذكاء الاصطناعي الإجرامي الناشئ. مراجعة الحوكمة وإعادة نشرها سنوياً. معدلات الخطأ تُحدَّث علناً. السباق لا ينتهي — والمراجعة أيضاً.

EXHIBIT 16SOURCES25:00

Authenticated Sources

Every claim — a primary authority · Full evidence compendium available as companion document
01
$12.5B documented internet crime losses (2023)
FBI IC3 Annual Report 2023 — ic3.gov
02
$3.94B investment fraud / pig-butchering
FBI IC3 Annual Report 2023
03
$25M Arup deepfake fraud (Feb 2024)
Hong Kong Police · BBC · Reuters
04
Ferrari voice-clone attempt (2024)
Bloomberg
05
$24.2B illicit cryptocurrency (2023)
Chainalysis 2024 Crypto Crime Report
06
WormGPT dark web LLM (Jul 2023)
SlashNext AI Security Report
07
Slovakia election deepfake audio (Sep 2023)
AFP Fact Check
08
Colonial Pipeline $4.4M ransom (2021)
US DOJ · FBI
09
State-actor AI influence ops (May 2024)
OpenAI Threat Intelligence Report
10
Operation Cronos — LockBit (Feb 2024)
NCA · FBI · Europol joint
11
INTERPOL Operation First Light (Jul 2024)
INTERPOL — interpol.int
12
1,040+ NPS in 120+ countries (2023)
UNODC World Drug Report 2023
13
50+ new NPS in Europe (2023)
EMCDDA European Drug Report 2023
14
Demographic bias — error rate differentials
NIST FRVT 2019 · pages.nist.gov/frvt
15
Governance standard — human-in-loop
INTERPOL–UNICRI Responsible AI for LE, 2023
16
Admissibility standards
ISO/IEC 17025 · Daubert v. Merrell Dow (1993) · GDPR Art. 46

مصادر موثّقة

كل ادعاء — جهة أولى موثوقة · وثيقة أدلة شاملة متاحة كمستند مرافق
01
١٢.٥ مليار دولار خسائر موثَّقة (٢٠٢٣)
FBI IC3 السنوي ٢٠٢٣ — ic3.gov
02
٣.٩٤ مليار احتيال استثماري
FBI IC3 السنوي ٢٠٢٣
03
احتيال Arup ٢٥ مليون دولار (فبراير ٢٠٢٤)
شرطة هونغ كونغ · BBC · Reuters
04
محاولة استنساخ صوت Ferrari (٢٠٢٤)
Bloomberg
05
٢٤.٢ مليار عملات مشفّرة غير مشروعة (٢٠٢٣)
Chainalysis 2024
06
WormGPT في الشبكة المظلمة (يوليو ٢٠٢٣)
SlashNext للأمن بالذكاء الاصطناعي
07
التزييف الصوتي الانتخابي السلوفاكي (سبتمبر ٢٠٢٣)
AFP Fact Check
08
Colonial Pipeline ٤.٤ مليون دولار (٢٠٢١)
DOJ الأمريكي · FBI
09
عمليات التأثير التابعة للدول (مايو ٢٠٢٤)
تقرير OpenAI للاستخبارات
10
عملية Cronos — LockBit (فبراير ٢٠٢٤)
NCA · FBI · Europol
11
عملية INTERPOL First Light (يوليو ٢٠٢٤)
الإنتربول — interpol.int
12
أكثر من ١٠٤٠ NPS في ١٢٠+ دولة (٢٠٢٣)
UNODC ٢٠٢٣
13
أكثر من ٥٠ NPS جديدة في أوروبا (٢٠٢٣)
EMCDDA ٢٠٢٣
14
التحيّز الديموغرافي — تباين معدلات الخطأ
NIST FRVT 2019 · pages.nist.gov/frvt
15
معيار الحوكمة — الإنسان في الحلقة
INTERPOL–UNICRI الذكاء الاصطناعي المسؤول، ٢٠٢٣
16
معايير قبول الأدلة
ISO/IEC 17025 · Daubert (1993) · GDPR المادة ٤٦
EXHIBIT 17CALL TO ACTION25:30

Join the Alliance

United Against Crime · One role, clearly defined · One standard, named

Governments

Deploy sovereign forensic and cyber AI. Invest in infrastructure-tiered national readiness. Align to ISO/IEC 17025 before any deployment.

Law enforcement

Validate the lead. Corroborate by accepted method. Bear professional responsibility. Testify. Human judgment — always. Per INTERPOL–UNICRI 2023.

Academia & research

Open models. Publish demographic error rates. Conduct adversarial testing. Hold us accountable to NIST FRVT. Keep us honest.

INTERPOL

Coordinate forensic, cyber, and financial intelligence within Constitution Article 2 scope. Build on First Light and Cronos. Peer-review the governance annually.

انضمّوا إلى التحالف

متحدون ضد الجريمة · دور واحد محدّد بوضوح · معيار واحد مُسمَّى

الحكومات

نشر ذكاء اصطناعي جنائي سيادي متدرّج. الاستثمار في الجاهزية الوطنية التدريجية. الالتزام بـ ISO/IEC 17025 قبل أي نشر.

جهات الإنفاذ

التحقّق من الدليل المبدئي. التحقّق بالطريقة المعتمدة. تحمّل المسؤولية المهنية. الشهادة. حكم الإنسان — دائماً. وفق INTERPOL–UNICRI 2023.

الأكاديميا والبحث

نماذج مفتوحة. نشر معدلات الخطأ الديموغرافية. إجراء الاختبار المعادي. إبقاؤنا مُلتزمين بـ NIST FRVT. إبقاؤنا أمناء.

الإنتربول

تنسيق الاستخبارات ضمن نطاق المادة ٢ من الميثاق. البناء على First Light وCronos. مراجعة الحوكمة سنوياً من الأقران.

EXHIBIT 18CLOSE26:00
Thank You
Capability under human command — never autonomy · Limits named — not hidden.
We will not always be ahead. We will not promise what we cannot deliver. But with the right data, the right algorithm, the right governance, and the right humility — we refuse to fall behind.
althowaikh@moi.gov.kw
شكراً لكم
قدرة تحت قيادة الإنسان — لا استقلالية · الحدود مُعلَنة — لا مخفاة.
لن نتقدّم دائماً. لن نعد بما لا نستطيع تقديمه. لكن مع البيانات الصحيحة والخوارزمية الصحيحة والحوكمة الصحيحة والتواضع الصحيح — نرفض أن نتأخّر.
althowaikh@moi.gov.kw
AI NARRATING
◈
SOVEREIGN AI
EXHIBIT 01
Evidence basis? Data sovereignty INTERPOL role False positives
◈ SOVEREIGN
I am the Sovereign Forensic AI system. Ask me anything about this framework — the evidence, the governance, the roadmap, or the technology. Every claim I make is sourced.
◈ ASK AI