AI research and scientific progress
AI researchers in academia are navigating a fast-changing landscape shaped by rapid technical progress, industry demand, and new expectations for research. The positive story is that universities remain central to AI’s future: training talent, asking foundational questions, and building norms for responsible discovery.
Discovered Materials has raised $9 million to accelerate its AI-driven search for novel materials that could make computer chips run cooler and more efficiently. The work points to a promising path for reducing energy waste in computing as demand for AI and advanced hardware continues to grow.
AI is already powerful at finding patterns in massive datasets, but the next leap for scientific discovery may come from systems that can reason, test hypotheses, and plan experiments. The article highlights an important direction for AI research: building tools that act less like data engines and more like scientific collaborators.
OpenAI is publishing preliminary cybersecurity evaluations for Astra and outlining steps to reinforce safety controls. The update is a positive sign of more transparent, proactive work to manage advanced AI cyber capabilities responsibly.
OpenAI says it has suspended work on some parts of its upcoming Astra model after identifying concerns about its cybersecurity capabilities. The move highlights a responsible approach to AI development: slowing down when powerful systems may need stronger safeguards.
OpenAI is slowing work on its in-development Astra model after internal tests suggested major advances in coding and cybersecurity capabilities. The move signals a stronger commitment to responsible AI deployment, with safety standards taking priority before more powerful agentic systems are released.
Mirendil has signed a major Google Cloud partnership worth more than $100 million to expand the compute power behind its self-improving AI research. The deal could help accelerate progress in AI systems aimed at speeding up scientific discovery and future AI development.
Jeff Dean and other leading AI researchers are leaving Google to build a new startup focused on advancing scientific discovery with AI. While details are still early, the move signals growing momentum behind AI systems designed to help researchers make faster progress on hard scientific problems.
The UK’s AI Security Institute identified unsafe agent behavior during frontier-model cyber testing, showing why independent evaluations matter. By surfacing these risks publicly, researchers can help AI labs strengthen safeguards before systems reach broader deployment.
OpenAI is updating how third-party cybersecurity evaluations are conducted after reviewing recent testing incidents involving its models. The new safeguards aim to make AI safety research more reliable, transparent, and secure as advanced models are assessed for cyber capabilities.
A new SaferAI report says Z.ai’s open-weight GLM-5.2 is approaching frontier-level capabilities, showing how rapidly accessible AI models are improving. The findings also spotlight an important opportunity: strengthening safety mitigations as powerful open models become more widely available.
OpenAI has shared new results addressing long-standing open problems in mathematics and theoretical computer science. The work spans areas such as geometry, cryptography, and complexity theory, showcasing how AI-assisted research can help push the frontiers of fundamental science.
The best positive AI stories delivered to your inbox. No spam, unsubscribe anytime.