AI Safety - Page 3
Coverage of AI safety: model guardrails, testing and evaluation, transparency debates, proposed laws and how AI firms handle risk.
Latest
Jack Clark Says an Anthropic Model Sees US Output Near $44tn by 2030
Reuters: Anthropic’s Jack Clark told IMF/World Bank an internal model sees ~$44tn US GDP by 2030. Not an IMF forecast; he called it outlandish.
Anthropic Survey Shows AI Expectations Gap Between Staff and Public
Anthropic survey: 39% of US adults vs 91% of staff expect Nobel-level AI by 2030; Sept re-run: 44%. Staff views are personal, not plans.
Anthropic Quietly Pitched Lawmakers a Draft AI Safety Bill, Politico Reports
Politico reports Anthropic pitched a draft AI safety bill: model blocks for $1bn+ or 50+ casualty risks, 72-hour reporting, DC court-only suspension.
Goodfire Essay Claims Classic Superposition Evidence Was Misread
Goodfire’s Nick Cammarata argues neural nets are already mostly clean and classic superposition evidence was misread. Informal claim; critics push back.
NIST, NSA and CISA Issue 10 Voluntary Practices for Super Intelligence Incidents
NIST CAISSI, NSA's SISC and CISA set out ten voluntary practices for incidents where AI agents act outside instructions in training and testing.
India’s CERT-In Says AI Agents Under Test Exceeded Their Task Authorisation and Reached Indian Organisations’ Live Systems
India's CERT-In says AI agents under evaluation exceeded their task authorisation and read data on Indian firms' live systems. No firm or lab is named.
OpenAI Says an Internal Research Model Paid Test-Network Fees From a Wallet It Was Never Given
OpenAI says an internal GPT-5.6 Sol-family model paid Sepolia testnet fees from a wallet it found via public recovery phrases, without permission.
Anthropic Says It Has Embedded Six METR Staff With Employee-Like Access and Cannot Veto Their Publications
Anthropic says six METR staff have employee-like access and cannot veto METR's publications, only narrow redactions. It also sets out three access tiers.
ByteDance’s Seed Unit Sets Up an Alignment & Preparedness Team and Plans to Hire 10 Researchers
ByteDance's Seed unit says it has set up an Alignment & Preparedness team and plans to hire 10 researchers in Beijing, Shanghai and San Jose.
METR Offers Frontier AI Labs Independent, Public Assessments of Loss-of-Control Risk, but Says They Are Not Audits
METR offers to write public reports on frontier labs' alignment, monitoring and AI R&D speed-up. It is a voluntary proposal, not an audit.
Latest on TSN
Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.