Eiropas profesionāļi sadarbojas modernā birojā

Apskats

Īss, Eiropai veltīts apskats par MI juridiskajā praksē. Pierādījumu līmeņi, jurisdikcijas konteksts un avotu saites katrā izdevumā.

Ieteicamais

Izdevums #50

The Board That Could Not Reconstruct What It Had Approved

2026. gada 22. augusts

In August a Latvian state agency lost the personal data of 1.2 million people and around 200,000 legal entities, drawn from eighteen years of payment records, under a five-year IT contract worth about EUR 9 million that ran through a chain of subcontractors. The supervisory board resigned on 19 August and the entire management board was gone by 20 August, because nobody at the top could say what the contract covered and what it left open. The same week computer use went generally available, seven in ten European large caps claimed board oversight of AI, and five per cent of S&P 500 directors were credited with AI experience.

Lasīt izdevumu

2026. gada 22. augusts

The Board That Could Not Reconstruct What It Had Approved

2026. gada 15. augusts

Seven and a Half Years, Down from Twelve

A manager said two things in one conversation: average tenure at her employer runs at about two years, and she would automate away twenty of her fifty people if the tools existed. The series says otherwise — US median tenure was 3.9 years in January 2024, higher than in 1983, 1987, 1998 and 2000, with voluntary quitting running below its pre-pandemic rate. What moved underneath the headline is the tail: a man aged 45 to 54 held 12.8 years with his employer in 1983 and 7.5 in 2024, the share of workers aged 25 and over with ten years or more fell from 33.3% to 30.2% in a decade, and an EDGAR search back to 2001 finds no US issuer that has ever disclosed the median tenure of its own workforce.

2026. gada 8. augusts

Ten Minutes of Help, and the Unassisted Score Falls

On 5 August a revised paper reported randomised controlled trials with 1,222 participants: AI assistance improved performance while it was present, and afterwards the same people scored significantly worse working alone and gave up sooner, with the effects appearing after roughly ten minutes of interaction. The day before, a Brooklyn judge sanctioned a lawyer whose nineteen submissions cited twenty-three fictitious cases forty-seven times, and who answered the challenge by asking the same generative systems to confirm their own output. This issue puts the two together against the controls that shipped the same week — pre-inference hooks, session budgets, a policy classifier — all of which govern the machine and leave untouched the human judgment that reads what it returns.

2026. gada 1. augusts

The Duty Got Lighter. The Liability Got Heavier.

On 27 July, Regulation (EU) 2026/1744 rewrote AI Act Article 4 from an obligation of result into an obligation of effort, while the deadlines, the enforcers and the penalty regime around it either held or hardened. Two days later Cornerstone Research reported that fifteen AI-related securities class actions — thirteen per cent of core filings in the first half of 2026 — carried seventy-three per cent of alleged investor losses, $385 billion of $529 billion. The statutory floor fell and the liability ceiling rose in the same fortnight, and the board question that follows is why the competence work was started in the first place.

2026. gada 25. jūlijs

The Repetition Loop Is Breaking

On 21 July the Thomson Reuters Institute put a name to the mechanism in its Future of Professionals 2026 law-firm survey: the repetition loop is breaking, and junior lawyers are being moved into advisory work ahead of the years of routine that built the instinct the advisory work requires. The next day Alphabet's quarter was read, priced and traded within hours, and on 24 July Regulation (EU) 2026/1744 handed every European AI programme sixteen months of extra runway. This issue sets the two clocks beside each other — one with a calendar, a room and an entitled challenger, the other with no dial at all — and follows the mechanism into Indeed's postings data, where entry-level roles are 4.5% of software-development openings.

2026. gada 18. jūlijs

A Regulator Just Asked Professionals to Show Their Working

On 15 July the Solicitors Regulation Authority closed a consultation that would ask a solicitor to record how they identified their own learning needs, keep the evidence for three years, and submit to a direction into prescribed learning where the regulator identifies a competence concern. The day before, a Texas bankruptcy judge ordered counsel who had filed a fabricated quotation into six hours of continuing education on generative AI in the courts — the same power, exercised on one lawyer. This issue explains why a record of self-identified needs inherits the blindness a Frontiers in Psychology study measured in AI users, and sets out the three controls — sampled re-judgments, seeded cases, and the volume nobody sees — that would make such a record evidential.

2026. gada 11. jūlijs

Thirty Per Cent of the Benchmark Was Broken

On 8 July OpenAI took apart SWE-Bench Pro, a coding benchmark it had recommended to the research community five months earlier, and found roughly 30% of its tasks faulty — hidden requirements, contradictory instructions, tests too strict to pass, grading that stopped short of the answer. Three days before that, Scientific American and Nature had assembled the evidence on what routine AI assistance does to a practitioner's own skill, including a Lancet study in which experienced endoscopists' unaided detection rate fell from 28.4% to 22.4%. This issue reads the two measurements together — one withdrawn by the people who made it, one taken by removing the machine and testing the human alone — and asks who maintains the instruments behind every capability figure in a board pack.

2026. gada 4. jūlijs

The BIS Reaches for the Canal Mania

The Bank for International Settlements used its Annual Economic Report to set the AI capital-expenditure boom beside canal mania, railway mania and the dotcom boom, recording that the five largest hyperscalers are set to spend over one trillion US dollars across 2025-26 in commitments that outpace their own earnings and free cash flow. In the same week the Council adopted the Digital Omnibus on AI, the GDPR track stalled when the Cyprus Presidency withdrew its compromise text, and the Commission adopted the revised ESRS - three instruments simplified, none of which asks what return a company expects on the money it is committing. This issue reads the report as a capital-allocation question rather than a market call: every board approving AI spend is underwriting a return, the board occupies the underwriter's position, and the practical obstacle is that few can see inside the deployment well enough to test the assumption.

Apskatu arhīvs