Corrections and retractions.
Published issues are not silently rewritten. Every correction, clarification and withdrawal is dated and listed here, and repeated on the issue itself.
Corrections and clarifications
- Issue 12 — The Nordic grid and the AI bill
- Clarification · 13 August 2026
Reserved grid capacity for data centres is reported at several stages of the connection queue, and figures differ accordingly: Statnett's Nordic Grid Development Perspective is the source of record here, and lower published totals reflect firm reservations rather than the full queue. The footnote now points at Statnett directly.
- Clarification · 13 August 2026
- Issue 11 — Pakistan's National AI Policy at one year
- Clarification · 13 August 2026
Disclosure: the author holds a personal advisory appointment connected to Pakistan's AI policy work. ImpactLab has no institutional partnership with the Government of Pakistan, and this issue is not written on its behalf.
- Clarification · 13 August 2026
- Issue 09 — When the biggest asset owner in the world screens AI with AI
- Correction · 13 August 2026
The machine-learning screening reported here is run by Norges Bank Investment Management's own ESG risk team, not by the Council on Ethics, which is a separate and independent body. The transparency point is unchanged; the attribution was not.
- Correction · 13 August 2026
- Issue 07 — Procurement is the underrated frontier of European AI policy
- Clarification · 13 August 2026
The five clauses set out here and the five contract conditions in Issue 03 are the same instrument, developed across both essays rather than two separate proposals.
- Clarification · 13 August 2026
- Issue 06 — The Norway–Pakistan analogue
- Correction · 13 August 2026
This issue originally dated the Norway–Pakistan Article 6 arrangement to April 2026. It was formalised in 2025. Corrected in the body.
- Clarification · 13 August 2026
ImpactLab has no institutional partnership with the Government of Pakistan. The connection referenced across this issue and Issue 11 is a personal advisory appointment held by the author, disclosed here so the reading is not mistaken for an organisational relationship.
- Correction · 13 August 2026
- Issue 04 — Norway exports energy. Virginia runs the models. The mismatch is the point.
- Correction · 13 August 2026
This issue originally named Cambridge as the place the models run. The intended referent was Northern Virginia, the largest concentration of frontier-model compute in the world. Corrected in the headline and body; the argument is unchanged.
- Clarification · 13 August 2026
Issue 08 restated this argument and has been withdrawn. This issue is the version of record.
- Correction · 13 August 2026
- Issue 03 — Diversity is downstream of vendor selection
- Clarification · 13 August 2026
The five contract conditions set out here and the five clauses in Issue 07 are the same instrument, developed across both essays. They are maintained as one document.
- Clarification · 13 August 2026
- Issue 02 — The Nordic model has trade-offs. Pretending otherwise hurts AI policy.
- Clarification · 13 August 2026
Issue 09 restated this argument and has been withdrawn. This issue is the version of record.
- Clarification · 13 August 2026
- Issue 01 — Why AI literacy is a policy instrument, not a training course
- Update · 13 August 2026
The enforcement dates named here have now passed: national market surveillance authorities took up the AI Act file on 2–3 August 2026. Norway's transposition had not entered into force on that date, so Norwegian deployers spent the first weeks of the enforcement era covered by the EU regime without a domestic supervisory authority named in force. Regulation (EU) 2026/1744, the Digital Omnibus on AI, has since amended the AI Act, and the Ministry of Digitalisation and Public Governance has announced a second consultation round on the Norwegian act to reflect it.
- Update · 13 August 2026
Withdrawn issues
Withdrawn · 16 April 2026
Issue 08 — Energy is the second AI policy nobody is reading
Restated the argument of Issue 04 with the same evidence base and no new claim. Withdrawn to keep one argument per issue.
Read the issue of recordWithdrawn · 7 May 2026
Issue 09 — What the Nordic model gets wrong about AI
Restated the argument of Issue 02 in shorter form and carried no footnotes. Withdrawn; Issue 02 is the version of record.
Read the issue of record