Sources: Google, Amazon, Facebook, Microsoft, and IBM form group to devise ethics for AI, and discuss its impact on jobs, transportation, warfare, and more
SAN FRANCISCO — For years, science-fiction moviemakers have been making us fear the bad things that artificially intelligent machines might do to their human creators.
Context & Ripple Effects
In September 2016 the five companies that dominated commercial AI research — Google, Amazon, Facebook, Microsoft, and IBM — were quietly building a shared ethics body, with jobs, transportation, and warfare on its agenda. Weeks later that effort was formalized as the Partnership on AI nonprofit, explicitly framed as best-practice sharing rather than government lobbying.
The arc since then has been a boom-bust in corporate AI governance: firms formalized internal ethics processes by 2018, their ethics chiefs later disclosed turning down client work like credit scoring and facial recognition (rejected client requests), and by 2023 many of those same responsible-AI teams were being cut even as a new generation of labs launched the Frontier Model Forum.
First-order effects
- The five founding members gain a joint venue for setting AI norms before regulators impose them, covering contested domains — employment displacement, autonomous transport, military uses — where each firm faces public scrutiny alone.
- IBM and Microsoft, which sell AI into enterprise and government markets, get an industry-endorsed ethics framework they can cite when customers question high-stakes deployments.
Second-order effects
- Rivals outside the founding five face pressure to join or match the group's standards or be cast as the unaccountable alternative, a dynamic that later drew new labs into successor bodies like the Frontier Model Forum.
- By writing 'best practices' as a private consortium rather than lobbying government, the members shape the baseline policymakers consult, keeping rule-setting inside the industry's own institutions.
Third-order effects
- If the pattern holds, AI governance becomes cyclical: coalitions form around each capability wave, codify principles, then get dismantled under cost pressure — leaving regulation to catch up after the fact, as the 2023 responsible-AI team cuts suggest.
- Self-policing by the largest vendors risks entrenching their standards as de facto industry rules, raising the barrier for smaller entrants who lack an ethics apparatus of comparable scale.
The trend: Corporate AI governance moves in waves — consortium formation, principle-setting, retrenchment — with each cycle transferring more norm-setting power to whoever still funds the ethics function.