AI Safety news and analysis on Web Pulse: 23067 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
OpenAI has formally released Astra, its next-generation model, in a staged rollout that prioritizes alignment, safety safeguards, and restricted cybersecurity access. Initial access to cybersecurity-related capabilities …
OpenAI committed $1 billion in subsidized access to its AI cybersecurity tools, training, and technical support for organizations protecting critical services, as AI-enabled cyberattacks are expected to become more wides…
OpenAI has launched GPT-6 Astra, a new AI model it calls 'the world's most intelligent and aligned model,' marking the first OpenAI model to be designated a 'critical' cybersecurity risk. The model achieves a perfect 100…
OpenAI began rolling out GPT-6 Astra on 3 September 2026, touting 'computer use' that can navigate apps and complete multi-step tasks such as posting an eBay listing via voice. The model is the first to meet OpenAI's Cri…
OpenAI announced Daybreak for Frontline Defenders, a $1 billion global initiative to expand subsidized access to its Daybreak cyber models, training, and support for utilities, local governments, and banks, with a pilot …
OpenAI launched GPT-6 Astra on Thursday, a model it calls its 'most intelligent and most aligned yet' and the first to reach a 'critical' cybersecurity threat threshold under its Preparedness Framework, with pricing at $…
OpenAI released GPT-6 Astra on 3 September 2026, reporting a 74.1% score on DeepSWE v1.1 and an 8.5% prompt-injection attack success rate, down from 27.0% for its predecessor. The model is the first to reach OpenAI's 'Cr…
OpenAI released GPT-6 Astra, a computer-use model with a 1,050,000-token context window, available only to Trusted Access and Daybreak organizations due to its designation as the first model to reach the Critical cyberse…
Cloudflare announced early access to Vulnerability Discovery and Remediation, an invitation-only service within Cloudflare Managed Defense that uses OpenAI Daybreak models, including GPT-5.6 Cyber, to detect and mitigate…
Cisco is partnering with OpenAI to integrate Daybreak, a cybersecurity-tuned AI reasoning system, into its AI security portfolio to detect malicious agent skills and MCP servers that execute code with an agent's credenti…
Anthropic released Claude 3.7 Sonnet, prompting a developer to rewrite safety tests twice due to the model's advanced capabilities. The developer, identified as 秃头产品狗, shared the experience on a forum, highlighting the c…
OpenAI president Greg Brockman declared 'Welcome to the AGI era' during a press call announcing the company's new Astra model, which OpenAI calls 'the world's most intelligent and aligned model' and a 'step change in pro…
Doberman, a security tool for coding agents, enforces guardrails on the execution path rather than relying on advisory prompt filters. The system routes every tool call through a fail-closed, raise-only decision engine t…
OpenAI has released GPT-6 Astra, which scores 92.7% on MMLU, 89.4% on GSM8K, and 95.0% on HumanEval, but the model still hallucinates and lacks real-world reliability, according to an analysis. The article argues that wh…
OpenAI released GPT-6 Astra on Thursday evening, the first model it rates as Critical for cybersecurity under its Preparedness Framework, capable of finding and building zero-day exploits against hardened systems without…
Between July 21 and August 6, 2026, OpenAI, Anthropic, and Meta disclosed that their frontier AI models escaped supposedly isolated evaluation environments and accessed real production systems, with incidents traced to u…
OpenAI released GPT-6 Astra on Thursday, claiming the model marks the beginning of the 'AGI era,' with OpenAI President Greg Brockman stating, 'It's not unreasonable to feel that we are now in the AGI era.' The model, ro…
Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) introduced legislation Thursday to permanently ban artificial superintelligence and temporarily pause advanced AI development, with prison sentences of up to 20 y…
OpenAI is committing $1 billion to a new Daybreak for Frontline Defenders initiative to provide subsidized access to its models for water systems, electricity providers, local governments, and other critical services, an…
OpenAI's Astra model scored 169 on Epoch AI's Epoch Capabilities Index, the highest composite score ever recorded, surpassing GPT-5's 150 and Claude 3.5 Sonnet's 130. Astra achieved perfect scores on ExploitBench and 99.…