Governance & Safety
Regulation, standards, safety, alignment, accountability, and institutional controls.
-
Governance & SafetyRead story
Who gets to say yes to AI data centers?
Ben Thompson argues that local approval could limit new AI data centers. That may help residents have a say, but it could also favor firms that already have sites.
Stratechery · Nov 4, 2026 -
Governance & SafetyRead story
State AI guidance names the risks. Who gets a voice?
A review of public school AI guidance from 35 states and Puerto Rico finds wide support for privacy and fairness. Few documents name students as contributors; none name parents.
arXiv cs.CY · Nov 5, 2026 -
Governance & SafetyRead story
OpenAI rewrote its rules. The hard cases remain.
OpenAI revised its Model Spec, a document about how its models should behave. Zvi Mowshowitz reads its five stated values and asks whether the rules hold up in hard cases. His critique is not a test of deployed models.
Don't Worry About the Vase · Nov 3, 2026 -
Governance & SafetyRead story
AI search can help you start. It cannot check the source for you.
WIRED's guide draws a line between finding a starting point and checking a fact. AI search may help you scan a broad topic. For an exact quote, a new update, or a high-stakes claim, open the original page. The guide offers advice, not a measured error rate for every search tool.
WIRED · Nov 2, 2026 -
Governance & SafetyRead story
Abusive image tools hid behind ordinary-looking AI apps
WIRED reports that researchers traced more than 400 linked apps and web domains in a network tied to abusive sexual deepfake tools. Apple said some approved apps later changed the websites they loaded. It removed 19 apps after WIRED contacted it. The report does not measure the whole app store.
WIRED · Nov 2, 2026 -
Governance & SafetyRead story
A monitoring test deserves more than a roundup
Zvi Mowshowitz relays a pilot test of how OpenAI watched AI systems that use tools. The account describes gaps in what could be seen and when people were alerted. OpenAI says it patched the relevant flaws. This is a newsletter account, not a fresh audit of the live system.
Don't Worry About the Vase · Oct 29, 2026 -
Governance & SafetyRead story
AI needs an action record, not just a human approval button
Srikumar and Leibowicz studied four systems that let AI use tools to do tasks. They say key records are missing: who ran the system, what permissions changed, and when a person stepped in. An approval is hard to audit if no one can check what happened next.
MIT Technology Review · Oct 29, 2026 -
Governance & SafetyRead story
A new AI grader may change the scores
Zhang and Yang tested three changes to an AI scoring tool using the same research proposals. After one change, 93% of scores had one value. The test did not use student work.
arXiv cs.CY · Oct 27, 2026 -
Governance & SafetyRead story
A correct AI draft can still miss the job
Work by new software students puts oversight beyond a right answer. Before approving an AI draft, ask whose job it serves and whether it fits that job.
arXiv · Oct 23, 2026 -
Governance & SafetyRead story
Blocking a data center may just move the burden
In a WIRED interview, Paris Marx warns that rejecting a data center may shift its power demand elsewhere. Residents need to see where the project might go next.
WIRED · Oct 25, 2026 -
Governance & SafetyRead story
A data center cannot reuse heat without a buyer
WIRED reports a proposed change to Germany's data-center heat rule. Reusing heat sounds simple, but a site needs someone nearby who can use it. The change was not yet law.
WIRED · Oct 21, 2026 -
Governance & SafetyRead story
AI helps find PHP flaws. Patching is still the test.
WIRED reports that experts used AI to find PHP flaws. Cloudflare saw millions of attempts to use those flaws. The PHP Foundation had seen no successful attacks when the story ran.
WIRED · Oct 24, 2026