Artificial intelligence can generate text, images and code while answering complex questions and switching between tasks. AGI, regarded as the next stage of this development, is expected to learn and reason across different fields and solve new problems; however, no such system exists yet. Google and DeepMind researchers have opened a discussion on the effects of AGI on the economy, working life, public policy and everyday life by establishing the DeepMind Institute (DMI), which also includes Demis Hassabis. Proposals include subjecting new models to 30 days of independent and confidential testing before they are released to the market, making evaluations mandatory in the future, and slowing the pace of development in a coordinated manner if risks increase. It was stated that the ideas do not represent Google’s official policy.
Why it matters
The focus of the discussion is shifting from progress in artificial intelligence capabilities as a purely technical issue to the possible consequences it could have for work arrangements, public decision-making and everyday life. The proposal for independent and confidential testing raises the need for oversight in evaluating new models that goes beyond developers’ own statements. The idea of jointly slowing the pace of development when risks increase leaves open the question of how coordinated action can be ensured among companies and researchers. The fact that the proposals are not Google’s official policy shows that they are a framework opened for discussion, not rules that have been implemented.
Background
Google is not a new name in the FikirPilot archive: we have published 39 news reports mentioning this name in the last 90 days; the latest is dated 17 September 2026.