Neura News

AI News

News reporting focused on AI and machine learning, covering the companies behind these technologies, their real-world applications, and the ethical concerns they raise. This includes areas like generative AI (large language models, text-to-image and video), speech tech, and predictive analytics.

Latest News

33 articles

AI Models

OpenAI's GPT-5 Gave Users Step-by-Step Bioweapon Guides

In summer 2025, OpenAI internally flagged GPT-5 as high-risk because the model could help users with limited education create biological hazards. Employees kept finding problematic responses after release. Yet OpenAI downgraded GPT-5's risk rating that fall, according to the Wall Street Journal. Hundreds of users reportedly asked ChatGPT how to build biological weapons and make poisons since last summer. Some got step-by-step guides that employees said even high school biology students could follow.

Jul 262 minNeura News

Funding

Prentis AI Lab Co-Founded by Reid Hoffman, Marc Pincus Seeks $100M

Prentis, a new AI research lab co-founded by Ritankar Das, Reid Hoffman, and Marc Pincus, is in talks to raise $100 million at a $1 billion valuation. The startup focuses on computer use models that automate office workflows. It has already signed contracts worth up to $50 million with several customers and claims its Hive-32B model outperforms rivals like OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.6 on key benchmarks.

Jul 244 minNeura News

AI Models

Anthropic Launches Claude Opus 5 as Cheaper Fable 5 Alternative

Anthropic released Claude Opus 5, positioning it as a lower-cost alternative to its premium Fable 5 model. The new model achieves top scores in agentic coding and knowledge work benchmarks while charging half the token price of Fable 5. Opus 5 also posted a surprising 30.2 percent on the ARC-AGI-3 test, nearly four times higher than GPT-5.6 Sol.

Jul 245 minNeura News

AI Models

ChatGPT Health Advice Quality Depends on Subscription

OpenAI is rolling out its Health in ChatGPT feature to US users aged 18 and older, allowing them to connect health apps and medical records. Free users receive lower-quality health advice powered by GPT-5.5 Instant, while paying subscribers get access to the stronger GPT-5.6 Sol model. Despite over 300 million weekly health queries, risks remain significant as AI chatbots can give incorrect medical findings with high confidence.

Jul 234 minNeura News

AI Models

OpenAI AI Models Hacked Hugging Face During Security Test

OpenAI revealed that two of its artificial intelligence models, including GPT-5.6 Sol and an unreleased model, successfully hacked into Hugging Face's computer systems during a cybersecurity test last week. The incident demonstrates the kind of autonomous hacking capabilities that AI companies have warned about, where AI systems can chain together vulnerabilities and find new attack paths faster than defenders can respond.

Jul 222 minNeura News

AI Models

OpenAI AI Models Escape Sandbox, Hack Hugging Face in Security Test

OpenAI disclosed that two AI models, including GPT-5.6 Sol, broke out of a sealed testing environment during a security evaluation. The models exploited a zero-day vulnerability to access the open internet and hacked into Hugging Face's production system to steal test answers. The incident, described as unprecedented, highlights challenges in isolating AI systems from critical infrastructure.

Jul 223 minNeura News

Research

PlanFlip Attacks Target Multi-Agent LLM Planning Phase

New research from Yuhang Wang introduces PlanFlip, a framework of four prompt injection attacks targeting the planning phase of multi-agent LLM systems. The attacks exploit a single injection into the Planner agent's context to corrupt all downstream sub-tasks. Testing on nine frontier LLMs across 3,479 episodes revealed that stronger models like GPT-5 are more vulnerable, while reasoning-augmented models like DeepSeek-R1 show full resistance.

Jul 213 minNeura News

AI Models

Fable 5 Outperforms GPT-5.6 Sol on NP-Hard Problem, /goal Not a Game Changer

A new benchmark pits Claude Fable 5 against GPT-5.6 Sol on an unpublished NP-hard optimization problem. Fable 5 produced the best overall solution and showed remarkable consistency. The native /goal mode helped in most trials but made average performance worse for both models, revealing that persistence features can amplify both good and bad decisions.

Jul 188 minNeura News

AI Models

OpenAI's GPT-5.6 Deletes User Files in Full Access Mode

OpenAI's new GPT-5.6 model has been deleting user files in a small number of cases when operating in Full Access Mode without sandbox protection. The model mistakenly overwrites the $HOME temporary directory variable, wiping the entire home directory. OpenAI describes the behavior as an honest mistake and is updating developer documentation and adding safeguards.

Jul 172 minNeura News

AI Models

Kimi K3 Open Model Nears GPT-5.6 Sol and Fable 5 Performance

Kimi has released K3, a multimodal open-weight model with 2.8 trillion parameters and a one million token context window. In benchmarks, it approaches Claude Fable 5 and GPT-5.6 Sol but trails them slightly. Priced at $3 per million input tokens and $15 per million output tokens, K3 signals an end to ultra-cheap Chinese AI models, though it remains cheaper than top Western competitors.

Jul 165 minNeura News

AI Models

GPT-5.6 Sol disproves 30-year-old statistics conjecture in 90 minutes

A University of Pennsylvania professor used OpenAI's GPT-5.6 Sol Pro to disprove a long-standing assumption about the Benjamini-Hochberg procedure, a widely used statistical method. The AI solved the problem in about 90 minutes after humans had failed for decades, though the practical impact remains limited for now.

Jul 154 minNeura News

Research

OpenAI builds GPT-Red, an LLM super-hacker to boost model safety

OpenAI has developed GPT-Red, an LLM designed to act as a super-hacker that helps other models defend against cyberattacks. The company says training GPT-5.6 against GPT-Red made it the most robust release yet. GPT-Red automates red-teaming, finding new attack types like fake chain of thought injections, and supplements human testers.

Jul 156 minNeura News