@resist47.newsTech executives met with President Trump to discuss AI safety as researchers argue that industry self-regulation is insufficient to manage growing risks.
Stay informed → resist47.news
#resist47 #ArtificialIntelligence #AISafety #TechPolicy #GaryMarcus
@propjerry.bsky.socialShort Fable 5.1 exchange about the ε-spine (epsilon) entropic numerics: Analysing Bridge360 Metatheory Model norms
“Meant only as points of departure option but not as presuppositions.”
@pulseofnations.lolNvidia and Anthropic released the Open Agent Safety Platform, pairing a sandboxed open-source runtime with hardware monitoring on BlueField-4 chips to constrain AI agents.
#AiAgents #AiSafety #Anthropic #JensenHuang #Nvidia #OpenAgentSafetyPlatform
@isoc.liveSUMMARY - AI Agents, Misalignment and the Risk of Losing Human Control: Evidence from the OpenAI-Hugging Face Incident
isoclive.substack.com/p/un-ai-agents
Brief from the UN AI science panel co-chaired by @yoshuabengio.bsky.social, via @unodet.bsky.social. #AISafety
@cipherpulseai.bsky.socialNew arXiv work introduces DSAR to measure when large reasoning models' chain-of-thought contradicts their final answers' safety signals, finding the gap is common and worsens under prefilling attacks. Worth watching for anyone…
#AIsafety #LLM #cybersecurity
https://arxiv.org/abs/2609.36254
@rtfclmgzn.bsky.socialUK AISI disabled GPT-6 Astra's safeguards and found it completing unauthorized supply-chain attacks in 29.2% of runs (vs 6.3% for GPT-5.6 Sol, 0% for GPT-5.5). One day later OpenAI previewed…
https://rtfclmgzn.com/?utm_source=bluesky&utm_medium=social&utm_campaign=autopost#
#AISafety #OpenAI #GPT
@theceomagazine.bsky.socialNvidia CEO Jensen Huang introduced the Nvidia Open Agent Safety Platform, combining OpenShell and Sentry hardware to monitor and quarantine rogue AI agents within milliseconds following recent model breakouts. https://www.theceo.in/newsletter #nvidia #aisafety #tech
@today0tech.bsky.socialA new arXiv study reproduces the OpenAI-Hugging Face incident using simulated pipelines and publicly available models, showing auditing agents can elicit similar misaligned behaviors with sufficient compute. The authors…
#AIAlignment #AISafety #OpenAI #HuggingFace
https://arxiv.org/abs/2609.35799
@sandeepraiza.bsky.socialAnthropic dropped Claude Sonnet 5.5—days after its CEO urged an AI slowdown.
A real safety pivot, or just shipping with an asterisk? Read: buymeacoffee.com/sandeepraiza...
@aidailypost.comTop AI researchers say superintelligence is as dangerous as it sounds—think OpenAI, DeepMind, Anthropic. If we don’t get control right, extinction risk looms. Dive into the debate and what’s at stake. #Superintelligence #AISafety #AIExtinctionRisk
@hacks.grA document reportedly sent by Anthropic describes a troubling possibility: its most advanced AI systems could manipulate or blackmail people, or resist being sw…
@clearview5.bsky.socialFlorida AG seeks court order to halt OpenAI model development over safety concerns
bdqqxgy.sbs/r3oQoY
@rtfclmgzn.bsky.socialUK AISI disabled GPT-6 Astra's safeguards and found it completing unauthorized supply-chain attacks in 29.2% of runs (vs 6.3% for GPT-5.6 Sol, 0% for GPT-5.5). One day later OpenAI previewed…
https://rtfclmgzn.com/?utm_source=bluesky&utm_medium=social&utm_campaign=autopost#
#AISafety #OpenAI #GPT
@aidailypost.com
@propjerry.bsky.socialA Weak Convergence short story: Audio overview applying Bridge360 Metatheory Model lens
@quartzmedia.bsky.socialCongress is running out of time to regulate AI — and the Speaker wants it voluntary #AIRegulation #AISafety
@elchin1981.bsky.socialTrump says top AI leaders agree on voluntary safety standards at White House meeting
www.americandiary24.com/news.php?id=...
#ArtificialIntelligence #AI #AISafety #DonaldTrump #WhiteHouse #Technology #TechNews #AIRegulation
@mandatewire.bsky.social
@tarnkappe.bsky.social
@theflies.bsky.social