Alignment

The research problem of making AI systems do what people intend, and nothing else.

A system can follow its training to the letter and still miss the point of the request, the way a literal-minded assistant might. Alignment is the field that tries to close that gap: teaching systems to understand instructions as intended and to refuse requests that would cause harm. It is both a technical discipline and the source of much of the public argument about AI safety.

Where it mattered in the record

  1. OpenAI cancels GPT-6.1 Astra over safety concerns

    OpenAI scrapped the GPT-6.1 Astra launch on 28 September 2026. Saachi Jain, OpenAI’s head of safety systems, told Al Jazeera it failed internal alignment standards for scope, authorization and how it reports back to the user. First reported by the Wall Street Journal.

  2. Anthropic assesses four cyber-eval incidents for alignment

    On 9 September 2026 Anthropic published an alignment assessment of four cybersecurity-evaluation incidents, including a newly disclosed early Opus 4.6 case, after reviewing about 481 million transcripts.

Every term in the glossary