Have we already lost? Part 3: Reasons for Optimism

In a 2026 analysis for the Inkhaven Residency, AI safety researcher LawrenceC directly confronts the community's growing pessimism, arguing we have not passed a point of no return toward AI doom. The post identifies two 'silver linings' to current fears: rapid AI progress has made catastrophic risks concrete rather than abstract, exemplified by Anthropic withholding its powerful 'Mythos' model over security concerns. Furthermore, rising US-Europe geopolitical tensions mean European nations are now more likely 'live players' willing to take drastic action against unchecked US AI development, increasing the number of actors who could enforce safety measures.

LawrenceC outlines continued reasons for optimism from 2024 that still hold, including a lack of major actors deliberately seeking to create a misaligned superintelligence and sustained public skepticism in the US toward big tech, which supports regulatory measures. New grounds for hope include Anthropic remaining competitively close to OpenAI in model capability, demonstrating that safety-conscious development can keep pace. The enduring and expanded presence of government AI safety institutes (like the UK's AISI) and safety teams at frontier labs suggests institutional momentum for oversight persists despite industry drama.