Google's global affairs chief just told regulators what they want to hear — that Big Tech can't be trusted to police itself.
The Summary
- James Manyika, Google SVP for Global Affairs, told Bloomberg that AI safety requires a "collective effort" beyond any single company
- A rare public admission from Big Tech that self-regulation isn't enough — likely positioning ahead of incoming frameworks
- The timing matters: This comes as the EU AI Act enforcement begins and Washington circles closer to comprehensive AI legislation
The Signal
James Manyika just handed regulators a gift-wrapped talking point. Google's SVP for Global Affairs said explicitly that AI safety "should not be left to any one company" and requires collective governance. For a company that spent the last decade fighting antitrust actions and resisting platform regulation, this marks a strategic shift.
The calculation is transparent. Google sees the regulatory train coming and wants a seat at the conductor's panel rather than a spot on the tracks. By advocating for industry-wide standards, they're betting that shared compliance costs hurt smaller competitors more than incumbents with legal armies and compliance infrastructure already in place.
"When Big Tech asks for regulation, they're usually asking for a moat."
But Manyika's framing reveals something else: the scaling problem is real. No single lab — not OpenAI, not Anthropic, not Google — has figured out alignment at the pace models are advancing. The compute requirements alone mean only a handful of players can train frontier models, but the safety challenges multiply faster than any one team can solve them.
Key realities driving this shift:
- Liability exposure from deployed AI systems is mounting faster than insurance markets can price
- China's centralized AI governance model is outpacing Western fragmentation
- The gap between model capabilities and safety guarantees widens with each generation
The "collective effort" Manyika references likely means something specific: a consortium model where major labs share safety research while keeping core IP locked down. Think CERN for AI safety, except funded by the same companies racing to ship product. Whether that structure can actually constrain competitive pressures or just provides regulatory air cover remains the open question.
The Implication
Watch what gets proposed, not what gets promised. If Google pushes for regulatory frameworks that require extensive compliance infrastructure, capital reserves, or government partnerships to operate AI systems, they're building barriers to entry while sounding responsible. The real test: will they advocate for regulations that constrain their own model deployment timelines, or just everyone else's?
For smaller AI companies and open source developers, this matters immediately. Industry-backed regulation tends to grandfather in existing players while creating compliance costs that crush new entrants. If you're building in the agent space, start mapping which proposed frameworks would kill your business model versus which you could actually work with.