Ressources

Ouvert en ce moment — des occasions d'agir avant la date limite

• How the FDA should evaluate AI in medical devices — public docket FDA-2026-N-7874
  États-Unis · Clôture 2026-10-19
• Colorado's rules for automated decisions and chatbots — public comment
  Colorado, États-Unis · Clôture 2026-10-26

Consultations closes

Comment ça marche, en bref

• What the AI Incident Database accepts
  The editor’s guide explains the database’s inclusion criteria; the database’s report form is the actual contact route.
• Finding a UK Parliament inquiry that is accepting evidence
  The inquiry finder is a discovery tool, not a recipient. Use it to locate a currently open call, then submit through that inquiry’s evidence page.
• A tracker is not always a reporting destination
  The MIT AI Incident Tracker classifies incidents obtained from the AI Incident Database; submit to the underlying database rather than the tracker.
• Inbound and outbound disclosure policies are not the same
  An outbound policy describes how a company reports flaws it finds elsewhere. It is not evidence that the company accepts reports through that page.
• Why third-party AI flaw disclosure needs real submission infrastructure
  A research proposal argues for standard reports, published testing rules, safe harbours and coordination across providers; it is not itself a disclosure route.
• The proposed US AI whistleblower law is not a reporting channel
  S.1792 was introduced legislation in the cited research, not enacted protection or a place to submit information.
• What happens to a consultation response
  When a regulator opens a consultation, the responses are usually published, read by the officials writing the rules, and sometimes answered one by one.
• Why an outside AI concern may go unheard
  Commentary about the field’s insularity can explain the problem, but it cannot receive a report. Prefer a recipient with a published duty or submission process.
• Why write to a committee, not only to your own representative
  Legislatures divide their work by committee. A message to the committee that hears AI bills reaches the people who decide what gets heard.
• Writing an AI flaw report that can be acted on
  Research on an open model bounty found that useful reports connect an observed output to a concrete deployment context and harm pathway.

Repris d'autres publications

Titres et liens uniquement. Nous ne copions jamais d'articles et n'intégrons jamais de vidéos. Le point de vue de chaque publication est indiqué, et l'équilibre entre les points de vue est publié.

• EFF Statement on California Governor's Executive Order on AI
  Electronic Frontier Foundation — Deeplinks · Civil liberties · Article · 2026-09-18
• Could AI really kill us all? Your questions, answered.
  MIT Technology Review — Artificial intelligence · Journalism · Article · 2026-09-18
• How Existential Fears Are Shaping the Debate Over AI
  AI Now Institute · Civil liberties · Rapport · 2026-09-17
• EU Online Age Checks Make Privacy Trade-Offs Unavoidable, CCIA Europe Warns
  Computer & Communications Industry Association (CCIA) · Industry lobby · Article · 2026-09-17
• Our framework for reporting model misalignment
  OpenAI — News · Industry · Article · 2026-09-16
• AISN #81: Anthropic Researcher’s Resignation Propels AI Risk into the Public Eye
  AI Safety Newsletter (Center for AI Safety) · Safety advocacy · Lettre d'information · 2026-09-15
• The week the AI freakout went mainstream
  Center for Security and Emerging Technology (Georgetown) · Academic · Rapport · 2026-09-12
• Congress must not waste the AI policy window
  Transformer · Journalism · Lettre d'information · 2026-09-11