Google DeepMind
Alphabet's AI research and frontier-model unit (Gemini), operating under Google's AI Principles and Frontier Safety Framework.
Google DeepMind builds Gemini and publishes one of the more detailed safety frameworks. In 2025 its parent dropped its no-weapons promise and moved into military contracts over staff objections.
Focus
- Safety evaluations
- Alignment research
- Transparency
Funding, as far as we know
Corporate (Alphabet) - https://abc.xyz/
What they've actually done
Google deleted its 2018 pledge not to build AI for weapons or norm-violating surveillance
4 February 2025
Published Frontier Safety Framework v3 adding harmful-manipulation and shutdown-resistance protocols; v3.1 April 2026
22 September 2025
Deployed Gemini to 3 million Pentagon personnel and negotiated a classified 'all lawful uses' deal, prompting a 580+ employee protest letter
2026-03 / 2026-04-27
Signed the EU GPAI Code of Practice
30 July 2025