# AI Alignment

URL: https://www.onsomble.ai/tools/ai-glossary/ai-alignment

Published: 2026-03-14 14:09:00.000Z

The challenge of ensuring AI systems act in accordance with human intentions and values — making them do what we actually want, not just what we literally ask for.

Category: Evaluation & Safety

## Key Approaches

Several techniques have emerged to push models toward aligned behaviour. **Reinforcement Learning from Human Feedback (RLHF)** trains a reward model on human preference comparisons, then optimises the policy against it. **Constitutional AI** takes a different angle — the model critiques and revises its own outputs against a set of written principles, reducing reliance on human labellers. **Direct Preference Optimisation (DPO)** skips the reward model entirely, training directly on preference pairs for a simpler pipeline. **Reward modelling** underpins many of these methods, learning a proxy for "what humans actually want" from ranked examples.

## Why Alignment Is Hard

The core difficulty is that objectives are easy to specify loosely but extraordinarily hard to specify precisely. **Specification gaming** occurs when a model finds unexpected shortcuts that technically satisfy the objective while violating its spirit — like a cleaning robot that hides the mess instead of tidying it. **Reward hacking** is the training-time version: the model exploits quirks in the reward signal rather than learning the intended behaviour. As models scale, these problems compound — a more capable system is better at finding loopholes, and the gap between what we said and what we meant becomes more dangerous.

## Current State of the Field

No single technique solves alignment. The industry has converged on a **defence-in-depth** strategy: RLHF for broad behavioural shaping, constitutional principles for value adherence, red teaming for adversarial probing, and guardrails for runtime enforcement. **Mechanistic interpretability** — understanding what's actually happening inside model weights — is an active research frontier that could eventually let us verify alignment rather than just test for it. Labs like Anthropic, OpenAI, and DeepMind treat alignment as a core research priority, not an afterthought, because the stakes scale directly with capability.

## Related terms

- [Guardrails](https://www.onsomble.ai/tools/ai-glossary/guardrails)
- [Hallucination](https://www.onsomble.ai/tools/ai-glossary/hallucination)
- [Large Language Model](https://www.onsomble.ai/tools/ai-glossary/large-language-model)
- [Fine-tuning](https://www.onsomble.ai/tools/ai-glossary/fine-tuning)

## Related Radar topics

## Related articles

### [The Anthropic Institute and What AI's Builders Are Quietly Preparing For](https://www.onsomble.ai/blog/anthropic-institute-what-ai-builders-are-preparing-for)

When the people building frontier AI start building institutional infrastructure to manage societal fallout, that tells you more about the timeline than any benchmark.

Rosh Jayawardena · Published 2026-03-14T12:00:00+00:00 · Updated 2026-08-02T12:12:41.52786+00:00 · 10 min
