IAS 38 Intangible Assets

IAS 38 applies to intangible assets that are not covered by another standard. For an in-house AI programme, that captures the trained model, a proprietary data asset that meets the definition of an intangible, and the internal development labour and compute that produce them. It does not cover goodwill, financial assets, or rights covered by other standards.

The standard distinguishes separately acquired intangibles, intangibles acquired in a business combination, and internally generated intangiblesIAS 38§25. An internal AI build sits in the third and most restrictive category.

IAS 38 requires an entity to classify the generation of an internal intangible into a research phase and a development phase. Expenditure on research is recognised as an expense when incurred, because at that stage the entity cannot demonstrate that an asset exists that will generate probable future economic benefits IAS 38§54.

Research phase → EXPENSEDevelopment phase → CONDITIONAL

An intangible arising from development is recognised only if the entity can demonstrate all six conditions in paragraph 57. If the entity cannot separate the research phase from the development phase, it treats the whole expenditure as if it were incurred in the research phase, that is, expensed.

Capitalisation in the development phase is permitted only when the entity can demonstrate, in summary: technical feasibility of completing the asset; intention to complete and use or sell it; ability to use or sell it; that it will generate probable future economic benefits; availability of adequate technical, financial and other resources to complete it; and the ability to measure the attributable expenditure reliably IAS 38§57. Each criterion is set out with evidenceable tests on the criteria page.

The cost of an internally generated intangible comprises the directly attributable costs necessary to create, produce and prepare the asset to operate as intended, incurred from the date the recognition criteria are first metIAS 38§66. Expenditure recognised as an expense in earlier periods cannot be reinstated as part of the assetIAS 38§71.

After recognition, an intangible with a finite useful life is amortised on a systematic basis over that life IAS 38§97. For fast-moving AI models, the useful life is typically short and must be justified.

IAS 38 sets a deliberately high bar for internally generated intangibles because the future benefit of early-stage work is uncertain. The practical consequence for AI programmes is that exploratory model building, most pre-training, and open-ended alignment work are expensed, while a targeted development-phase build that meets the criteria can be capitalised from the point the criteria are met.