Orion-1 outperforms Claude and GPT-6 on cyber defense
The Decoder: AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap
Artemis Security's new Orion-1 model is purpose-built for cyber defense and beats Claude Opus 5.5 and GPT-6 Sol on the company's own benchmark. The model fills a growing gap as general-purpose frontier models lock down their cyber capabilities, blocking defenders in the process. The article AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap appeared first on The Decoder.
Get smart on it
A startup has launched a specialized AI model designed to investigate and respond to cyberattacks on corporate networks independently, outperforming general-purpose AI models on security-focused tasks. The model matters because major AI providers are restricting their general-purpose models' cybersecurity capabilities to prevent misuse by attackers, which inadvertently blocks legitimate security defenders from using those same tools for analysis and response. Security teams face a gap where the guardrails meant to stop hackers also prevent defenders from leveraging powerful AI systems, creating a need for purpose-built alternatives. The startup built its own benchmark to test security decision-making rather than relying on existing tests, and the specialized model scored significantly higher on attack reconstruction tasks compared to general-purpose competitors.
Explainer: How to Actually Read an AI Benchmark (Without Being Fooled)
Relevant courses
Want to go deeper than the news? Explore live, cohort-based AI courses taught by practitioners.
Browse AI courses on MavenRelated news
Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed
Google Deepmind's AlphaProof Nexus solved dozens of open math problems, including nine Erdős problems two of which were unsolved for 56 years, per a study in Science. The publication comes as OpenAI faces a boycott for dumping AI proofs on GitHub without peer review. Deepmind had its results reviewed before publishing. The article Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed appeared first on The Decoder.
‘Pure insanity’: Mathematicians will need years to make sense of OpenAI’s latest drop
"Staggering." "Overwhelming." "Unprecedented." "Surreal." "Pure insanity." Those were among the descriptions more than three dozen mathematicians reached for in conversations with The Verge as they tried to make sense of the flood of mathematical results OpenAI abruptly dropped on the field this week. Amid the awe, excitement, and uncertainty over the sheer scale of the deluge was a deep-seated anxiety over what it all means - and what comes next. For all their different react
Open, peer-reviewed measurement for today’s AI buyers: MLPerf in the EU’s AI Gigafactories tender, and Endpoints v1.0 opens for submissions
As AI adoption is growing across the economy, AI buyers are asking a more complex set of questions about AI systems than simply a chip’s speed. They want to know what performance a deployed service will deliver, what it will cost for given workloads, and how much energy it will use. MLPerf has been the standard-bearer for measuring AI system performance since 2018; it’s evolving to address these more complex questions for every AI buyer. First, out of Europe: MLPerf has been chos
Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB
Z.ai just released GLM-Legal. GLM-Legal is a legal AI model from Z.ai, built on GLM-5.3-Flash and trained for agentic legal work. It targets contract review, evidence analysis and compliance research. The key result is a 19.2% all-pass rate on Harvey’s Legal Agent Benchmark (LAB). That is nearly 3x its base model’s 6.7%. For developers, the catch is access: the API is application-only for now. TL;DR Size: Built on GLM-5.3-Flash, a 320B-total, 18B-active MoE with a 1M-token co
Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots
Sakana AI's MASS method lets one language model improve itself by solving tasks, judging results, and refining workflows, reaching a win rate of nearly 70 percent against the base model after two rounds. But when the model misses the same errors across all three roles, it can mistake them for progress and reinforce them during training. The article Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots appeared first on The Dec
AI agents overstate their results and remain far from autonomous research, study finds
Epoch AI and Anthropic independently found the same thing: current AI models like GPT-5.6 Sol and Claude Fable 5 can run experiments but lack scientific self-criticism and genuine creative thinking. At best, Sol reached 15 percent of the human reference score, and even that came from methods researchers already knew. The models' biggest weakness is still their inability to critically question their own results. The article AI agents overstate their results and remain far from autonomous