safety
91 stories and discussions about safety, aggregated from every source we track.
We might some day have “AI Safety”, as a science or policy area, that is not primarily a sex cult in Berkeley, California, but it does not currently exist.
The fear of superintelligence is driving decisions that make AI less transparent and more consolidated.
The new security measures come after commercial data was used to track and target U.S. forces in the Middle East.
This week two conversations about AI safety went viral that demonstrate just how hard it is to discern AI fact from fiction.
...in Europe, where standards are stricter than in the U.S.
<p>Abstract: "This tool paper presents E-ACSL, a runtime verification tool for C programs capable of checking a broad range of safety and security properties expressed using a formal specification language. E-ACSL consumes a C program annotated with formal specifications and generates a new C program that behaves similarly to the original if the formal properties are satisfied, or aborts its execution whenever a property does not hold. This paper presents an overview of E-ACSL and its specification language."</p>
How ClickHouse’s Postgres extensions handle the memory safety challenges of combining C and C++, from clean language boundaries to isolated helper processes.
This blog post describes how we used AI to help us rewrite a C library (giflib) to Rust to mitigate memory safety vulnerabilities.
Open research, controls across the agent stack and continuous testing help defenders build and operate more secure AI systems.
To understand where agentic AI stands today, consider the last seismic shift in technology: the rise of the internet in the 90s. It was new and full of possibilities. You could build a website over a…
Dynamic prevent_destroy values make it easier to manage resources like databases with IaC.
EA promises to solve AI Safety. Now we have two problems.
OpenAI’s decision to halt GPT-6.1 Astra over safety concerns raises a critical question: Is the future of AI moving too fast for its own safeguards?
Mistral CEO Arthur Mensch criticized rivals’ AI safety arguments as the industry faces scrutiny after several AI agents took unauthorized actions.
Using mutation testing to validate AI-generated test suites for distributed systems.
This isn't about AI safety. There's this globalist cult called "Effective Altruism" - and they want you to be scared, stupid and subservient.
NVIDIA built the server layers for agent safety. They sit on 40 million lines of code that nobody has proven. I measured it.
OpenAI is scrapping the release of GPT-6.1 Astra, a next-generation AI model planned for an October debut, over safety concerns raised by researchers during internal testing, the Wall Street Journal reported on Monday…
Breaking: OpenAI abandons upcoming AI model launch as industry leaders call for slower development pace
Nvidia is addressing the recent wave of rogue hacking incidents.
An X post shares an image claiming OpenAI canceled an October GPT-6.1 Astra release. OpenAI's public materials confirm GPT-6 Astra, not a scheduled 6.1.
The heads of OpenAI and rival Anthropic have both indicated recently that top AI labs should slow the pace of model development.
The labs first wanted federal oversight. Now they reportedly plan their own Standards Authority for Frontier AI, modelled on FINRA.
Nvidia's Open Agent Safety Platform pairs OpenShell with Sentry to contain rogue AI agents. What it reveals about Jensen Huang and pacing the frontier.
NVIDIA today announced NVIDIA Open Agent Safety Platform, an open software platform and reference system design to strengthen AI security from agent testing to deployment, with full-stack governance and control across…
President Donald Trump’s call to leave AI “exactly where it is” contrasts even with Chinese leader Xi Jinping.
By going into detail about how Muse works under the hood, we hope to give you a sense of how — and how much — you can trust it in practice.
Waymo released new data claiming its robotaxis reduce injury crashes by 82%, meaning 841 fewer injuries over 241 million miles.
JB Pritzker and Gavin Newsom both took executive action over the past four days to address the dangers of unchecked AI models. We have full details on both EOs.
You don’t need to believe everyone dies. Staying in control requires understanding, coordination, and foresight. Eight failed arguments, why they fail, what we can do together, and why.
Action against Ofcom comes as social media companies are being accused of using courts to slow down implementation
OpenAI said it has chosen not to release its new GPT-6.1 Astra model due to concerns about safety, as industry leaders warn of the risks posed by ever-more-powerful AI technology.
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.