#AI Security

Articles about AI Security — exploring patterns, best practices, and real-world implementations in production systems.

35 posts tagged with ai security. ← Blog home

#A/B Testing (3)#A2A (14)#ADK (8)#AG-UI (6)#AI (9)#AI Adoption (9)#AI Agents (355)#AI Architecture (22)#AI Cost (10)#AI Cost Optimization (8)#AI Engineering (316)#AI Evaluation (9)#AI Financial Engineering (14)#AI Gateway (8)#AI Governance (43)#AI Red Teaming (9)#AI Research (17)#AI Safety (12)#AI Security (35)#AI Strategy (8)#AI in Production (12)#AML (3)#API Design (17)#API Security (9)#APIs (69)#AWS (17)#Accountability (8)#Accounting (9)#Agent Memory (8)#Agent Skills (11)#Agentic AI (25)#Agentic Commerce (12)#Agentic Engineering (12)#Agentic RAG (8)#Agents (4)#Amazon Bedrock (8)#Analytics (3)#Architecture (45)#Attention (3)#Audit (4)#Authentication (11)#Authorization (3)#Automation (13)#Azure (11)#Azure AI Foundry (9)#Backend Engineering (322)#Benchmarks (3)#Best Practices (5)#BigQuery (6)#Billing (8)#Build Systems (4)#Business Finance (8)#Business Strategy (55)#C (8)#CI/CD (23)#Caching (11)#Capital Markets (14)#Card Payments (12)#Cards (13)#Career (30)#Causal Inference (8)#Cellular Automata (8)#Change Management (3)#Checkpointing (4)#Claude Code (8)#Cloud (5)#Cloud Architecture (3)#Cloud Native (10)#Code Review (9)#Coding Agents (20)#Collaboration (6)#Communication (9)#Complexity (8)#Compliance (54)#Computer Networking (9)#Computer Science (40)#Computer Vision (15)#Computer-Use Agents (8)#Concurrency (42)#Consulting (26)#Containers (10)#Context Engineering (19)#Conversational AI (8)#Cost Optimisation (5)#Cost Optimization (3)#Credit (16)#Credit Risk (16)#CrewAI (8)#Crypto (12)#Cryptocurrency (12)#Cryptography (9)#Custody (9)#DSPy (8)#Data (14)#Data Engineering (25)#Data Governance (3)#Data Science (8)#Data Structures (10)#Databases (48)#Deep Learning (20)#Deployment (5)#Design (3)#Design Patterns (10)#DevOps (24)#DevSecOps (22)#Developer Experience (7)#Developer Productivity (15)#Developer Tools (5)#Differential Privacy (3)#Diffusion Models (10)#Discovery (5)#Distributed Systems (104)#Documentation (3)#Edge AI (8)#Embeddings (21)#Emotional Intelligence (8)#Energy (8)#Engineering (14)#Engineering Culture (3)#Engineering Judgment (7)#Engineering Practices (21)#Enterprise AI (8)#Error Handling (4)#Evaluation (79)#Event-Driven (9)#Event-Driven Architecture (8)#Explainability (3)#Explainable AI (8)#FREE-AI (8)#FX (5)#Feature Engineering (3)#Feature Flags (3)#Feedback (4)#FinOps (23)#FinTech (6)#Financial AI (14)#Financial Engineering (8)#Financial Systems (133)#Fine-Tuning (12)#Fintech (134)#Flutter (8)#Forecasting (10)#Foreign Exchange (5)#Forward Deployed Engineer (43)#Forward Deployed Engineering (8)#Fraud (10)#Fraud Detection (8)#Function Tools (5)#Functional Programming (3)#Fundamentals (23)#Fundraising (8)#GCP (5)#GPU (8)#GUI Automation (8)#Gemma (4)#Generative AI (13)#Git (8)#Go (232)#Go-to-Market (31)#Google ADK (36)#Governance (61)#Granite (6)#Graph Neural Networks (8)#GraphQL (3)#Growth (3)#Guardrails (35)#HIPAA (3)#HTTP (3)#Harness Engineering (8)#Hiring (9)#Hugging Face (8)#Human-in-the-Loop (10)#IBM watsonx (8)#Idempotency (3)#Identity (11)#Image Generation (10)#Inference (8)#Information Retrieval (8)#Integration (3)#Integrations (8)#Intellectual Property (8)#Interfaces (3)#Interpretability (8)#JavaScript (8)#KYC (11)#KYC and AML (12)#Kafka (11)#Knowledge Graphs (11)#Kubernetes (17)#LLM (5)#LLM Inference (8)#LLM Infrastructure (8)#LLM-as-Judge (3)#LLMOps (9)#LLMs (197)#LangChain (8)#LangGraph (11)#Leadership (33)#Ledger (12)#Legal (8)#Lending (14)#Linux (9)#LlamaIndex (8)#Load Balancing (3)#MCP (25)#MLOps (39)#Machine Learning (119)#Marketing (16)#Markets (4)#Mathematical Finance (8)#Memory (16)#Memory Management (5)#Metrics (7)#Microservices (3)#Microsoft Agent Framework (150)#Middleware (6)#Migration (9)#Migrations (8)#Mixture of Experts (5)#Model Risk (3)#Monetization (9)#Monitoring (5)#Monorepo (3)#Multi-Agent (10)#Multi-Agent AI (14)#Multi-Agent Systems (76)#Multimodal (3)#Multimodal AI (11)#NIM (5)#NVIDIA (8)#Networking (3)#OAuth (3)#OWASP (7)#Observability (53)#On-Device AI (8)#Ontology (8)#Open Source (7)#OpenAPI (3)#OpenTelemetry (5)#Operating Systems (9)#Operations (14)#Opinion (6)#Orchestration (10)#Organizational Design (8)#Payment Rails (16)#Payments (55)#People (8)#Performance (56)#Personalization (9)#Platform Engineering (17)#Policy as Code (3)#PreSales (8)#Pricing (4)#Privacy (14)#Privacy Engineering (3)#Process (4)#Product (30)#Product Management (9)#Production (22)#Programming (10)#Programming Languages (48)#Prompt Engineering (86)#Prompt Injection (16)#Proof of Concept (3)#Protocol Buffers (3)#Protocols (9)#Providers (4)#Pydantic AI (8)#Python (142)#Quality (3)#Quantitative Finance (22)#RAG (72)#RBI (3)#REST (5)#Rails (16)#Real-Time (8)#Reasoning Models (8)#Recommender Systems (8)#Reconciliation (5)#RegTech (8)#Regulation (10)#Reinforcement Learning (9)#Reliability (68)#Resilience (4)#Responsible AI (5)#Retrieval (3)#Risk (22)#Risk Management (11)#Rust (32)#SLSA (3)#SRE (23)#SaaS (9)#Sales (9)#Sales Engineering (8)#Scalability (6)#Search (10)#Security (103)#Security Engineering (8)#Self-Evolving Agents (24)#Semantic Web (8)#Sessions (3)#Settlement (9)#Soft Skills (8)#Software (3)#Software Architecture (36)#Software Delivery (9)#Software Engineering (152)#Spanner (4)#Specification Engineering (3)#Speech (8)#Startups (30)#Statistics (8)#Strands (8)#Strategy (6)#Stream Processing (9)#Streaming (38)#Structured Output (4)#Supply Chain Security (10)#Sustainability (9)#System Design (32)#Systems Programming (56)#Technical Sales (6)#Testing (79)#Threat Modeling (3)#Time Series (9)#Tool Use (23)#Tooling (11)#Tools (4)#Trading (8)#Transformers (3)#Treasury (6)#Trust (4)#Type System (3)#Type Systems (12)#TypeScript (8)#Validation (3)#Valuation (3)#Vector Databases (23)#Vector Search (12)#Venture Capital (8)#Version Control (9)#Versioning (7)#Voice AI (9)#Web Development (6)#Webhooks (9)#Workflow (4)#Workflows (14)#eBPF (8)#gRPC (13)#smolagents (8)
Pratik Dhanave · ·5 min read

Closing the Loops: Composition, Governance, and What Self-Evolution Is For

The series mapped three loci and a feedback axis as if systems politely stayed in one square of the grid. They don't — the direction of the field is composition: artifact loops whose lessons distill into the harness, harness accumulation that consolidates into weights, each locus covering the others' weaknesses. This closing post adds the final axis — where the loop closes — then assembles the full map, reads off the governance that composition demands, and ends on the only success metric that resists gaming: what the system actually delivers.

The third axis — loop closure, from within-task to online-continuous, with gating rising to match — completes the map. The frontier is composed stacks: artifact loops feeding harness distillation feeding rare, gated weight consolidation under monotone feedback strength. Governance reads off the three axes; success is measured in delivered artifacts. Closes the series.

Pratik Dhanave · ·4 min read

The Feedback Axis: Every Self-Evolution Failure Is a Feedback Failure

Walk back through the first six posts and a pattern repeats with suspicious regularity: the artifact loop is only as good as its evaluator, consensus training launders whatever bias the vote shares, self-play drifts without an anchor, test-time adaptation mis-fires when its proxy objective lies. Different loci, same lesson. This post makes the lesson the subject: feedback is the taxonomy's second axis, failure modes sort by feedback type rather than by locus — and the engineering discipline of self-evolving systems is mostly the engineering of feedback.

Feedback is the taxonomy's second axis: five types on a strength ladder — formal, mechanical, environmental, judged, self-consistent — each with a signature failure, all governed by Goodhart at type-specific pressure thresholds. The universal architecture is the cascade: cheap filters, strong gates, inter-stage agreement as the drift alarm, humans as calibrators.

Pratik Dhanave · ·5 min read

What an AI Production Readiness Audit Actually Covers

"Is our AI system ready for production?" is a question most teams answer with a feeling. An audit replaces the feeling with evidence: a structured pass over the six dimensions that decide whether an AI system survives real users, real data, real cost, and a real security review. Here's exactly what mine covers and what you walk away with — useful even if you run it yourself.

'Is our AI ready for production?' answered with evidence instead of a feeling: the six dimensions an audit covers — success criteria, evaluation, grounding, cost and latency at volume, security, reliability and ownership — with the red flags for each, and what you walk away with. Useful even if you run it yourself.

Pratik Dhanave · ·5 min read

Preparing Security Teams for AI Adoption

Security teams meet enterprise AI twice: first as the thing employees are already using without asking, then as the thing the business formally wants to deploy. A security function prepared for both becomes the reason AI adoption is safe and fast; an unprepared one becomes either the department of no — which drives usage underground — or a rubber stamp that discovers the risks in an incident report. This post is the adoption program's view of AI security: what's genuinely new, and how to get ahead of it.

Security meets AI twice: as shadow AI employees already use, then as the systems the business wants to deploy. Treat shadow AI as demand signal, learn the genuinely new surface — prompts as exfiltration, injection via content, agents as non-human identities, fluent wrongness — and build the paved road with risk-tiered review so the safe path is the fast path.

Pratik Dhanave · ·6 min read

Agentic Software Engineering in Go: Security & Safety

Prompt injection defense, sandbox execution, and RBAC

The security engine in Go: input sanitization, prompt-injection defense, command sandboxing, role-based access control, and audit logging — plus incident response and the security metrics dashboard. Agents are untrusted-input processors with execution privileges; this post treats them that way.

Pratik Dhanave · ·3 min read

Shipping AI: Production, Security & Reliability

Proving value on a prototype is the halfway point. The other half — and the harder one — is making it survive production: evaluated, secured, compliant, reliable, and owned by the customer's team. This post maps the shipping theme of the field guide, where most AI initiatives actually stall.

Proving value on a prototype is the halfway point; the harder half is making it survive production — evaluated, secured, compliant, reliable, and owned by the customer's team. Deployment, evaluation, AI security, sovereignty and compliance, reliability, and handover, plus the new AI topics (agent security, operationalized governance, sovereign AI, evals as infrastructure).

Pratik Dhanave · ·6 min read

Evaluating and Operating Guardrails

A defense you haven't tested is a hope, not a control. The final discipline of LLM security is treating your guardrails as a system to be measured, attacked, and monitored continuously — because the threat evolves, your application changes, and a defense that worked last quarter can silently rot. This closing post covers red-teaming your own system, operating it in production, and the honest state of the art.

A defense you haven't tested is a hope, not a control. The final discipline of LLM security is treating your guardrails as a system to be measured, attacked, and monitored continuously — because the threat evolves, your application changes, and a defense that worked last quarter can silently rot. Red-teaming, production monitoring, and the honest state of the art.

Pratik Dhanave · ·6 min read

Guardrails in Practice

"Guardrails" is the umbrella term for the runtime checks that sit around a model and screen what goes in and comes out — content classifiers, moderation models, topic and format validators, PII detectors. They're a real and useful layer, distinct from the architectural defenses, and they come with their own design rules: layer them, fail safe, and never mistake them for a wall.

"Guardrails" is the umbrella term for the runtime checks around a model — content classifiers, moderation models, topic and format validators, PII detectors. They're a real and useful layer, distinct from architectural defenses, with their own design rules: layer them, fail safe, and never mistake them for a wall. This post covers building the layered defense in practice.

Pratik Dhanave · ·6 min read

Treating Model Output as Untrusted

Injection defense usually focuses on what goes into the model. But an equally dangerous class of bug lives on the way out: whatever the model produces gets passed to another system — a browser, a shell, a database, another service — that trusts it. If the model can be made to emit a malicious payload, and your code renders or executes it, the injection escapes the model and lands in your infrastructure.

Injection defense usually focuses on what goes into the model, but an equally dangerous class of bug lives on the way out: whatever the model produces gets passed to another system that trusts it. If the model can be made to emit a malicious payload and your code renders or executes it, the injection escapes the model and lands in your infrastructure — XSS, SQLi, SSRF, exfiltration.

Pratik Dhanave · ·6 min read

Architecture and Least Privilege: The Real Defense

Everything before this post raised the probability barrier against injection. This post lowers the impact — and impact is what actually protects you. The load-bearing defense against prompt injection isn't a prompt or a filter; it's an architecture where a fully-hijacked model still can't do anything catastrophic, because it was never granted the power to.

Everything before this raised the probability barrier against injection; this post lowers the impact — and impact is what actually protects you. The load-bearing defense isn't a prompt or a filter but an architecture where a fully-hijacked model still can't do anything catastrophic: least privilege, the confused-deputy trap, human-in-the-loop, and the dual-LLM pattern.

Pratik Dhanave · ·5 min read

Prompt Hardening and Its Limits

Between filtering the input and re-architecting the system sits a tempting middle ground: make the prompt itself more resistant. Delimiters, spotlighting, instruction placement, and defensive system prompts all raise the cost of an attack. None of them close the hole — because they are all still text in the one channel the attacker also writes to — but used well they meaningfully shift the odds.

Between filtering input and re-architecting the system sits a tempting middle ground: make the prompt itself more resistant. Delimiters, spotlighting, instruction placement, and defensive system prompts all raise the cost of an attack — but none close the hole, because they're all still text in the one channel the attacker also writes to. The value is knowing exactly how much they buy.

Pratik Dhanave · ·5 min read

Input Defenses and Their Limits

The first instinct when facing prompt injection is to inspect the input and block the bad stuff. It's a reasonable layer — but a treacherous one, because it creates a feeling of safety far larger than the protection it provides. This post covers the input-side defenses that are genuinely worth having, and draws a hard line around what they can and cannot do.

The first instinct against injection is to inspect the input and block the bad stuff. It's a reasonable layer but a treacherous one — it creates a feeling of safety far larger than the protection it provides. This post covers the input defenses genuinely worth having, and draws a hard line around what they can't do (starting with: indirect injection bypasses them entirely).

Pratik Dhanave · ·6 min read

A Taxonomy of Injection and Jailbreak Attacks

You can't defend against what you can't categorize. Prompt-based attacks come in a few structurally distinct shapes — direct injection, indirect injection through content the model reads, and jailbreaks that target the model's safety training — and each demands a different defense. This post maps the attack surface so the rest of the series can defend it systematically.

You can't defend what you can't categorize. Prompt-based attacks come in structurally distinct shapes — direct injection, indirect injection through content the model reads, and jailbreaks targeting the model's safety training — and each demands a different defense. This post maps the attack surface so the rest of the series can defend it systematically.

Pratik Dhanave · ·6 min read

Why Prompt Injection Is Unsolved

Prompt injection is the defining security problem of LLM applications, and — unlike SQL injection, which it superficially resembles — it has no clean fix. The reason is structural: a language model reads instructions and data through the same channel, and cannot reliably tell which is which. This opening post explains why that makes injection fundamentally hard, and reframes the goal from "prevent it" to "contain the blast radius."

Prompt injection is the defining security problem of LLM applications, and unlike SQL injection it has no clean fix — because a model reads instructions and data through the same channel and can't reliably tell them apart. This opening post explains why that makes injection fundamentally hard, and reframes the goal from prevent to contain the blast radius.

Pratik Dhanave · ·5 min read

Limits, Risks, and Open Problems

The frontier is genuinely exciting and genuinely oversold, and telling the difference matters — so this closing post is an honest accounting of what self-evolving agents cannot yet do, what can go wrong, and what remains unsolved.

The frontier is genuinely exciting and genuinely oversold. An honest accounting of what self-evolving agents cannot yet do, what can go wrong, and what remains unsolved.

Pratik Dhanave · ·5 min read

Evaluating Open-Ended Improvement

Every frontier method is a search, and a search is only as good as the evaluator that ranks its candidates — so at the frontier, evaluation stops being a measurement and becomes the single most dangerous component in the system.

Every frontier method is a search, and a search is only as good as the evaluator that ranks its candidates — so at the frontier, evaluation becomes the single most dangerous component in the system.

Pratik Dhanave · ·6 min read

Shipping MCP Securely to Production

An MCP server can run code and see context on the model's behalf, which makes it powerful and dangerous in equal measure — this is how to deploy one without handing attackers the keys.

An MCP server can run code and see context on the model's behalf. Authentication, prompt-injection and tool-poisoning risks, human-in-the-loop, sandboxing, and a production checklist.

Pratik Dhanave · ·5 min read

Security and Adversarial Robustness

AI systems add attack surface that traditional security does not cover — the model, its prompts, its retrieved context, and its tools are all attackable — and the only way to know you're defended is to threat-model the whole surface and prove it with red-teaming.

AI adds attack surface conventional security misses — model, prompts, retrieved context, and tools are all attackable. Phase 7: threat-model the whole surface and prove it with red-teaming.

Pratik Dhanave · ·5 min read

Evaluating Evolution — and Keeping It Safe

A system that changes itself can improve itself right off a cliff, so the evaluation and guardrails are not an afterthought to self-evolving agents — they are the thing that makes them safe to run at all.

A system that changes itself can improve right off a cliff. Measuring evolution honestly, reward hacking, drift and collapse, and the guardrails that keep it safe.

Pratik Dhanave · ·8 min read

Building an AI Red-Team Program

A single pre-launch red-team decays the moment your model, prompt, or tools change — turning adversarial testing into a sustained program is what keeps an AI system safe past day one.

The capstone: making red-teaming a sustained program — the remediation loop where findings become regression tests, blending automated/manual/external modes, rules of engagement, ship-blocking severity thresholds, governance evidence, and culture.

Pratik Dhanave · ·14 min read

AI Red-Teaming and Testing

The series finale: how to proactively find AI security failures before attackers do — turning injection, leakage, and excessive-agency risks into a repeatable adversarial test suite that runs in CI, measuring attack success honestly, and standing up incident response for the day a control fails.

Find AI security failures before attackers do: adversarial testing of the whole system, a test taxonomy mapped to the series, automated tooling (PyRIT, garak), a CI security-regression suite that fails the build, honest attack-success-rate measurement, and AI incident response.

Pratik Dhanave · ·12 min read

Measuring and Scoring Red-Team Results

Turning red-team attacks into metrics you can act on and track over time — attack success rate, coverage, severity, and trend — plus the honest limits of what any of those numbers can tell you.

Turning attacks into metrics: attack success rate and why it's subtle, scoring success (rule/classifier/LLM-judge with its biases), coverage across the taxonomy, severity weighting, tracking trends per model/prompt version, and honest reporting of residual risk.

Pratik Dhanave · ·13 min read

Securing the AI Pipeline

Part seven of the AI Security Engineering series: DevSecOps for AI systems — securing the secrets, network, supply chain, prompts, and CI/CD gates that surround the model, so a hardened model doesn't sit inside a soft pipeline.

DevSecOps for AI: secrets in a manager not code, least-privilege runtime identities (no ambient prod creds for agents), egress control, rate/spend limits against model DoS, supply-chain verification in CI, prompts-as-code, and a security-eval gate that fails the build.

Pratik Dhanave · ·13 min read

Automated Red-Teaming and Tooling

Scaling red-teaming beyond manual probing — the building blocks of an automated harness (seed library, mutation, orchestrator, scorer), LLM-driven adaptive attackers, the real tools by role (PyRIT, garak, promptfoo, Giskard), and wiring it all into CI as a repeatable gate.

Scaling red-teaming: the harness building blocks (attack seeds, mutation, orchestrator, scorer), adaptive LLM-driven attackers, the real tools by role (PyRIT, garak, promptfoo, Giskard), and integrating an automated red-team gate into CI.

Pratik Dhanave · ·14 min read

Guardrails and Content Safety

The defensive layer that screens what goes into a model and what comes out — input rails, output rails, topical rails, and groundedness checks — plus the real tooling ecosystem and a vendor-neutral Python pipeline that wraps a model call and knows how to refuse.

The defensive layer that screens inputs and outputs: input/output/topical/groundedness rails, the real ecosystem (Llama Guard, Granite Guardian, NeMo Guardrails, Guardrails AI, Presidio, hosted moderation), and a vendor-neutral guardrail pipeline — with honest false-positive/negative trade-offs.

Pratik Dhanave · ·13 min read

Red-Teaming Agents and RAG Systems

Why agents and retrieval turn a prompt injection into real-world action, how to red-team the highest-risk AI surface with benign canaries, and the least-privilege controls that shrink an attacker's blast radius.

The highest-risk modern surface: indirect injection via RAG/tools, tool abuse and excessive-agency exploitation, memory poisoning, multi-step attacks, and data-exfiltration channels — with a canary methodology and least privilege as the primary control.

Pratik Dhanave · ·14 min read

RAG and Supply-Chain Security

Part five of the AI Security Engineering series: the two OWASP LLM risks that live in the plumbing around the model — the documents your agent retrieves and the models, datasets, and dependencies it is built from — and the Python patterns that treat both as untrusted until proven otherwise.

Securing the components around the model: RAG as an injection and poisoning vector (treat retrieved content as hostile, enforce provenance and per-user authz) and supply-chain risk (safetensors over pickle, model provenance, pinned and vetted dependencies and tools).

Pratik Dhanave · ·13 min read

Data and Model Attacks

A defender's tour of the attacks that target the model and its data — prompt and context extraction, training-data memorization, membership inference and model inversion, model stealing, poisoning and backdoors, and evasion — with what a red-teamer tests and what actually stops each one.

Attacks on the model and its data: system-prompt and training-data extraction, membership inference and model inversion, model stealing, and poisoning/backdoors — what to test and defend, and the honest risk difference between using hosted models and training your own.

Pratik Dhanave · ·12 min read

Insecure Output Handling and Excessive Agency

Two tightly-linked OWASP LLM risks that turn a clever prompt injection into real-world damage — and the Python patterns that shrink the blast radius: treat model output as untrusted input, and give agents the least agency they can get away with.

Two OWASP risks that turn an injection into damage: insecure output handling (model output is untrusted input — never eval/shell/SQL it unescaped) and excessive agency (least-privilege tools, allow-lists, human approval for irreversible actions, audit logs).

Pratik Dhanave · ·12 min read

Prompt Injection and Jailbreaks in Depth

A defender's field guide to the injection and jailbreak techniques a red-teamer probes for — the taxonomy, why each one works, and how to turn it into a re-runnable test suite that maps every passed test to a concrete fix.

A deeper, test-focused look at injection and jailbreak families — direct vs indirect (the RAG/agent threat), role-play, obfuscation, many-shot, multi-turn, cross-lingual — plus a red-team methodology: build a probe suite, mutate, test input and retrieval paths, measure, re-test.

Pratik Dhanave · ·13 min read

Data Security and Privacy

Part three of the AI Security Engineering series: protecting the data that flows through an LLM system — how sensitive information leaks out of prompts, logs, and retrieval, and the engineering controls (redaction, data minimization, per-user retrieval authz, residency choices) that actually stop it.

Protecting data in LLM systems: sensitive-information disclosure, PII in prompts and logs (your observability can be the leak), redaction with Presidio, data residency, and per-user access control on the retrieval layer so RAG doesn't leak across tenants.

Pratik Dhanave · ·11 min read

Threat Modeling and the AI Attack Taxonomy

Before you attack an AI system you need a map of it: the components an adversary can influence, the trust boundaries between them, and a taxonomy that sorts attacks by goal and stage so your red-teaming is systematic instead of a grab-bag of the attacks that happen to trend that week.

Mapping the AI attack surface so red-teaming is systematic: threat-modeling the components (model, prompts, training/RAG data, tools, guardrails) and an attack taxonomy by goal (integrity/availability/privacy/abuse) and stage (training vs inference), aligned to NIST AI 100-2 and ATLAS.

Pratik Dhanave · ·14 min read

Prompt Injection and Jailbreaks

Why the #1 risk on the OWASP LLM Top 10 has no clean fix — the model can't tell your instructions apart from the text it reads — and the defense-in-depth pattern that actually shrinks the blast radius.

The #1 LLM risk: direct vs indirect prompt injection (the latter is the real threat for RAG and agents), jailbreak families, and honestly-rated defense-in-depth — delimiting untrusted data, least-privilege tools, human-in-the-loop, and detection's limits.

Pratik Dhanave · ·11 min read

What AI Red Teaming Is

The opening post of a hands-on series for builders: what it means to systematically stress-test an AI system — model, prompts, retrieval, tools, and guardrails — to surface its failures before adversaries or ordinary users do, how that differs from pentesting and robustness testing, and the frameworks and ethics that keep the work rigorous.

The opener to a defensive AI red-teaming series: stress-testing AI systems (models + prompts + RAG + tools + guardrails) to find failures before adversaries do — how it differs from pentesting, the frameworks (OWASP GenAI, MITRE ATLAS, NIST AI 100-2), and ethical rules of engagement.

Pratik Dhanave · ·11 min read

The AI Security Landscape

Why LLM and agent applications open a genuinely new attack surface, the mental models to reason about it (OWASP Top 10 for LLM Applications, MITRE ATLAS, NIST AI RMF), and how to threat-model an AI system before you write a line of defensive code.

The opener to a series on securing LLM and agent systems: why AI apps have a new attack surface (instructions and data share one channel, tools grant real power), the OWASP Top 10 for LLM Applications, MITRE ATLAS, and how to threat-model an AI system.

All posts on this site are written by Pratik Dhanave, an Agentic AI Architect with 7+ years building production distributed systems, multi-agent AI platforms, and cloud-native infrastructure. About the author → Each article includes working code, architecture diagrams, and references to the specific frameworks and standards discussed. Browse the blog or explore related topics using the tag cloud above.