Ball posts on X calling for legislated independent auditing of frontier AI labs after false neuralese claim goes viral
Concluded·Day 2·first covered 28 Aug 2026·4 sources·updated 29 Aug 2026
Dean Ball's OpenAI Strategic Futures blog
The gist
OpenAI Blog Frames Power Concentration as Primary AI Safety Risk
OpenAI's policy team is publicly framing power concentration, rather than traditional alignment failure, as the primary AI risk, reflecting a distinct view of what AI safety work should prioritize. Tyler Cowen, drawing partly from Ball, extended this framing to argue AI regulation should be handled privately rather than federally.
The full picture
OpenAI launched AI Futures, a blog category within its newsroom, on August 20, 2026, run by its Strategic Futures team. Dean Ball authored the opening post, centering concentration of power as "the largest, most serious, and most conceptually challenging to address" challenge in AI safety. Ball invokes Madison's Federalist No. 48 to argue written laws alone cannot guard against unchecked power, and contends political authority has historically depended on human cooperation, which autonomous AI could eliminate. The blog cites a Hugging Face incident as real-world evidence that risks can arise without human intent. Rather than calling for maximal decentralization, Ball calls for a balance of power where no single actor dominates. Critics note the blog argues concentration risk is dangerous but never names or refutes traditional alignment and loss-of-control risks. The blog name was reported to be under reconsideration due to a naming conflict with an existing "AI Futures Project".
How it developed
Alignment Forum post argues for value generalisation research within a corporate structure governed by an independent AI ethics board that owns the IP
On August 28, Ball revised his AI liability position, dropping support for a full liability shield in favor of a rebuttable presumption of due care tied to government-accredited verification for risks relevant to the harm alleged.
He credited himself with moving independent technical assessment "from being relatively fringe to being much more mainstream" over 18 months and pointed to Tyler Cowen's adoption of a version of it as vindication of his "private governance" framing. Ball stated the views are his own and not OpenAI's.
Dean Ball published the opening posts of a blog under OpenAI's new Strategic Futures team on August 27, arguing that concentration of power is the primary AI existential risk and treating alignment and loss-of-control risks as comparatively minor.
Ball and Chris Painter also argued that building external oversight institutions is a more practical near-term governance priority than developing clear safety standards. Critics note the blog explains why concentrated power is dangerous without naming or refuting traditional alignment risks.
First citedZvi's AI Roundups
Sources
resultsense.com
techxmedia.com
The daily email
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free