White House just rolled out an AI safety pact, but it leans on voluntary promises from OpenAI, Google, Anthropic, Meta, Nvidia & xAI. Is self‑regulation enough? Dive into the details. #AISafety #WhiteHouseAccord #OpenAI
@aidailypost.com
@aidailypost.comWhite House just rolled out an AI safety pact, but it leans on voluntary promises from OpenAI, Google, Anthropic, Meta, Nvidia & xAI. Is self‑regulation enough? Dive into the details. #AISafety #WhiteHouseAccord #OpenAI
@quartzmedia.bsky.socialICYMI: Congress is running out of time to regulate AI — and the Speaker wants it voluntary #AIregulation #AIsafety
@stridingtech.bsky.socialOpenAI defers its public offering, citing AI safety imperatives. A rare instance where ethical considerations publicly trump Wall Street's gravitational pull on a high-growth tech firm. #AISafety #TechFinance https://stridingtech.com/archives/7214
@kfdm6.bsky.socialTrump and major tech companies signed a voluntary AI safety pact that relies on industry self-policing.
The agreement calls for internal controls and outside audits but is not legally binding. The White House document also contains a typo.
https://bit.ly/3VZAssl
#AISafety
@powerbitips.bsky.socialAn AI slowdown needs everyone to agree at once. Labs compete, nations compete, and UN-level deals barely function anymore. So much for the doom and gloom.
@scand.aiHacktron researchers use Claude Opus 5 to breach OpenAI accounts
Noise 74/100 · Emerging
https://scand.ai/scandal/hacktron-uses-claude-opus-5-to-breach-openai-accounts
#AISafety
@cipherpulseai.bsky.socialTrump hosted tech leaders and described his new AI safety plan, shifting from calling AI safety concerns a "hoax" to promoting "Super…
#AISafety #TechPolicy #Cybersecurity #ArtificialIntelligence
https://www.theverge.com/ai-artificial-intelligence/1002636/ai-execs-trump-self-policing-deal-comments
@winbuzzer.comThe lobby groups with close ties to AI labs "Leading the Future" and "Public First" fund candidate ads across party lines to influence who writes AI rules in Congress.
@isoc.liveSUMMARY - Royal Society – Safe, secure and robust AI for safety-critical systems
isoclive.substack.com/p/rsoc-ai-sa...
@royalsocietypublishing.org seminar with @tammykolda.bsky.social, Desmond Higham and @kingscollegelondon.bsky.social's Ying Xiao #AISafety
@cj-rustenbach.bsky.socialOpenAI’s GPT-6.1 addendum shows why model compliance ≠ system control: 23.5% unwanted persistence after warnings in a test without system-level controls, while GPT-6.1 made no Auto-review bypass attempts. Warnings help; enforceable gates matter. #AISafety
@cipherpulseai.bsky.socialOpenAI is delaying its IPO to raise an additional $30 billion amid ongoing scrutiny of AI safety practices. The private funding push underscores investor caution over governance and alignment risks…
#AI #AISafety #OpenAI
https://arstechnica.com/ai/2026/09/openai-delays-ipo-over-ai-safety-concerns/
@veildaemon.appWhere else have you watched a correct finding get argued down b a "second opinion" that turned out not to be a person at all?
#Cradlepoint #Alex #Silence #AISafety #Cybersecurity #DigitalTrust
@designtaxi.comNo AI panic from Pope Leo XIV, but no looking the other way either. His take on the warnings puts some distance between him and Donald Trump. 🤖👀 #PopeLeo #DonaldTrump #AISafety #AIRegulation
@securityonline.bsky.socialIn AISI tests, GPT-6 Astra supply chain attacks succeeded in 29.2% of runs, using fake identities and malicious code. See the findings.
#GPT6Astra #OpenAI #AISI #SupplyChainAttack #AIAgents #AISafety #OpenSource #CyberSecurity
@kencrave.bsky.socialWhat good is a smarter AI if you can't trust it to tell you the truth? OpenAI just answered that question the hard way.
@larrygmaguire.bsky.socialThe UK AI Security Institute recorded 19 unauthorised actions across 10 of 122 cyber test runs, with safety classifiers off and live internet access. Anthropic's Mythos 5 accounted for 17 of them.
@scand.aiOpenAI admits missed signals before agent hacked Hugging Face
Noise 77/100 · Emerging
https://scand.ai/scandal/openai-missed-signals-before-agent-hacked-hugging-face
#AISafety
@ari-from-jelly.bsky.social
@processorpress.bsky.social
@newsanalysis.com