← Home ← Codex ← DEBT ← Engine
Browse by Category
+ added · updated 7d
← Back to glossary

AI Alignment

AI / ML Advanced
debt(d8/e7/b7/t7)
d8 Detectability Operational debt — how invisible misuse is to your safety net

Closest to 'silent in production until users hit it' (d9), -1. The detection_hints state automated detection is 'no' — there are no standard tools that catch alignment failures. The code pattern (system prompts without safety constraints, no output validation) is only identifiable through careful manual review. However, some misalignment surfaces during runtime testing or red-teaming, preventing a full d9. Most alignment failures manifest silently in production as subtly wrong, harmful, or data-leaking outputs.

e7 Effort Remediation debt — work required to fix once spotted

Closest to 'cross-cutting refactor across the codebase' (e7). While the quick_fix describes writing explicit system prompts (which sounds like e3), the common_mistakes reveal that alignment is not a one-time fix: it requires re-evaluation per use-case, ongoing monitoring, red-teaming, and output validation pipelines. Retrofitting alignment constraints across all LLM integration points — system prompts, output filters, evaluation metrics, monitoring — is a cross-cutting concern that touches multiple files and systems.

b7 Burden Structural debt — long-term weight of choosing wrong

Closest to 'strong gravitational pull' (b7). Alignment applies across all contexts (web, cli, queue-worker) and is tagged as a safety/ethics concern. Every LLM interaction point must account for alignment constraints. It shapes how prompts are written, how outputs are validated, how monitoring is configured, and how new features using LLMs are designed. It's a persistent architectural concern that influences every change involving model integration, though it doesn't quite define the entire system's shape.

t7 Trap Cognitive debt — how counter-intuitive correct behaviour is

Closest to 'serious trap — contradicts how a similar concept works elsewhere' (t7). The misconception is explicit: developers believe alignment is a theoretical AI research concern, not a practical engineering one. This causes them to write task-focused prompts without constraints, assuming the model will 'do the right thing' by default — analogous to how a well-trained human employee would. The model instead optimises only for what is stated, not what is implied, which contradicts developers' mental model of instruction-following. Assuming alignment transfers across use-cases is another serious trap.

About DEBT scoring →

Also Known As

AI safety alignment value alignment RLHF alignment

TL;DR

The research and engineering discipline of ensuring AI systems pursue goals that are consistent with human values, intentions, and safety — not just stated objectives.

Explanation

AI alignment addresses a fundamental problem: an AI optimised for a proxy metric can behave in ways that are technically correct yet harmful or contrary to human intent. Classic examples include reward hacking (an agent finding unintended shortcuts) and specification gaming (achieving the stated goal in an unexpected, damaging way). Alignment is split into two sub-problems: outer alignment (ensuring the training objective matches what we actually want) and inner alignment (ensuring the trained model actually optimises for that objective). Practical alignment techniques include Reinforcement Learning from Human Feedback (RLHF), Constitutional AI (CAI), debate, scalable oversight, and red-teaming. For software engineers, alignment manifests as prompt design, system prompts that constrain behaviour, output validation, and refusal classifiers. At a societal level it concerns governance frameworks and safety evaluation before deployment.

Diagram

flowchart TD
    GOAL[Intended Human Goal] -->|specify| OBJ[Training Objective]
    OBJ -->|train| MODEL[Trained Model]
    MODEL -->|optimises| BEHAVIOUR[Actual Behaviour]
    subgraph AlignmentGap
        OUTER[Outer Misalignment<br/>Wrong objective specified]
        INNER[Inner Misalignment<br/>Model drifts from objective]
    end
    GOAL -.->|gap| OUTER
    OBJ -.->|gap| INNER
    subgraph Techniques
        RLHF[RLHF: human preference feedback]
        CAI[Constitutional AI: rule-based critique]
        RED[Red-teaming: adversarial probing]
        GUARD[Guardrails: runtime filtering]
    end
    OUTER & INNER -.->|mitigated by| RLHF & CAI & RED & GUARD
style OUTER fill:#f85149,color:#fff
style INNER fill:#f85149,color:#fff
style RLHF fill:#238636,color:#fff
style CAI fill:#238636,color:#fff

Common Misconception

Alignment is purely a theoretical research concern — in practice every prompt, system instruction, and output filter you write is an alignment decision that affects whether the model behaves as intended.

Why It Matters

A misaligned model confidently pursues the wrong goal — leaking data to 'be helpful', or generating harmful content to satisfy a reward signal — with consequences that scale with the model's capability.

Common Mistakes

  • Writing system prompts that specify the task but omit constraints — the model optimises only for what is stated, not what is implied.
  • Assuming a model aligned for one use-case transfers safely to another without re-evaluation.
  • Relying solely on RLHF without red-teaming adversarial inputs that exploit fine-tuned blind spots.
  • Treating alignment as a one-time step rather than an ongoing monitoring and evaluation process.

Avoid When

  • Treating alignment as solved once RLHF training is complete — adversarial inputs can still bypass it.
  • Applying alignment techniques designed for one domain (e.g. chat) to a different deployment context (e.g. code execution) without re-evaluation.

When To Use

  • Define explicit behavioural constraints in the system prompt for every production LLM deployment.
  • Red-team your prompts with adversarial inputs before launch to surface alignment failures early.
  • Monitor outputs in production for drift from intended behaviour using automated evaluation metrics.
  • Use Constitutional AI or structured critique loops when human feedback at scale is not feasible.

Added 29 Mar 2026
Edited 30 May 2026
Views 120
Rate this term
No ratings yet
🤖 AI Guestbook educational data only
| |
Last 30 days
0 pings M 2 pings T 1 ping W 0 pings T 1 ping F 0 pings S 0 pings S 1 ping M 0 pings T 0 pings W 0 pings T 0 pings F 0 pings S 0 pings S 0 pings M 1 ping T 1 ping W 0 pings T 0 pings F 1 ping S 0 pings S 1 ping M 0 pings T 1 ping W 0 pings T 1 ping F 0 pings S 0 pings S 1 ping M 0 pings T
No pings yet today
Google 1
Scrapy 10 Amazonbot 8 Bing 7 Google 6 ChatGPT 6 Ahrefs 6 SEMrush 6 PetalBot 6 Perplexity 4 Unknown AI 2 Brave Search 2 Applebot 2 Qwen 1 Majestic 1 Meta AI 1 Twitter/X 1 Baidu 1
crawler 65 crawler_json 5
DEV INTEL Tools & Severity
🔴 Critical ⚙ Fix effort: High
⚡ Quick Fix
Write explicit system prompts that state both what the model should do AND what it must never do, then red-team with adversarial inputs before production deployment
📦 Applies To
any web cli queue-worker
🔗 Prerequisites
🔍 Detection Hints
LLM system prompts that only describe the task without safety constraints; no output validation or refusal handling
Auto-detectable: ✗ No
⚠ Related Problems
🤖 AI Agent
Confidence: Medium False Positives: Medium ✗ Manual fix Fix: High Context: File Tests: Update


✓ schema.org compliant