OpenAI has reportedly slowed the development of its upcoming Astra model due to security concerns arising from its advanced capabilities. This decision was made in response to the model's potential for misuse, particularly in sensitive areas.

The slowdown is attributed to Astra's significant advancements in agentic coding and cybersecurity, which have reached a 'Critical' threshold under OpenAI's Preparedness Framework. This move underscores OpenAI's commitment to the safe development and deployment of advanced AI systems.

👉 Read the full article on TechCrunch

OpenAI announced that its upcoming Astra model has demonstrated significant advancements in agentic coding and cybersecurity, reaching a 'Critical' threshold under its Preparedness Framework. This evaluation suggests the model could potentially identify and develop zero-day exploits of all severity levels or devise and execute end-to-end novel cyberattack strategies without human intervention.

In response, OpenAI is implementing stricter security controls and pausing certain internal activities to ensure the safe and secure further development of Astra. These measures include isolated testing environments, restricted network and tool access, enhanced model weight protections, and additional monitoring capabilities.

The company emphasizes collaboration with government agencies and AI safety organizations to test the model's capabilities, asserting that advanced cyber-capable models should aid defenders in identifying vulnerabilities before attackers. This commitment to transparency and responsible deployment is a crucial aspect of OpenAI's safety approach.

👉 Read the full article on OpenAI

DeepMind has reportedly developed a new AI model capable of predicting hurricanes earlier than any other existing system. This breakthrough has the potential to revolutionize weather forecasting and disaster preparedness.

The AI model is believed to leverage advanced machine learning techniques to analyze complex meteorological data patterns, predicting hurricane paths and intensities more quickly and accurately than traditional physics-based models. This capability could provide communities with crucial additional time to prepare, enabling more effective measures to protect lives and property.

👉 Read the full article on Wired