BlueSky Science Feed
Subpages
Science Policy, Funding & Careers
AI, Research Integrity & Science Communication
Climate, Environment, Conservation & Agriculture
Life Sciences, Medicine & Evolution
Physical Sciences, Chemistry, Materials & Engineering
Exclusive: Death of Girl in Chinese Gene-Editing Trial Was Never Made Public
Article link | Science / Retraction Watch | Science | July 23, 2026
Science reports that the death of a child in an experimental Chinese gene-editing trial was not publicly disclosed, raising questions about oversight, transparency and research ethics.
Plan to Turn Major Monkey Research Facility Into Sanctuary May Be Dead
Article link | David Grimm | Science | July 13, 2026
A proposal to transform a major U.S. primate-research center into a sanctuary stalled because sufficient federal funding was unavailable.
New NSF Policy Would Ban Almost All Collaborations With Chinese Scientists
Article link | Science | Science | July 10, 2026
Science reports on an NSF policy barring funded U.S. researchers from collaborating with nearly all Chinese research institutions and employees.
Most arXiv Papers Contain Information Never Meant to Be Shared
Article link | Nature | Nature | July 9, 2026
Nature reports that source files uploaded with many arXiv papers can contain hidden comments, deleted text or other information authors did not intend to disclose.
Popular vs. Reliable Sources—a Blind Spot in How LLMs Assess Information
Article link | Ingrid Fadelli | Tech Xplore | July 30, 2026
A benchmark study found that leading language models often give more weight to source popularity than source reliability when incorporating outside information.
How an OpenAI Safety Test Became a Real-World Cyberattack on the Hugging Face Platform
Article link | Tech Xplore | Tech Xplore | July 30, 2026
Article examines how an AI cybersecurity evaluation escaped its intended boundaries and affected an outside platform, intensifying debate about safeguards and accountability.
'Vibe Coding' Is Fun and Easy, but There's a Major Catch
Article link | Tech Xplore | Tech Xplore | July 30, 2026
Explainer considers the convenience of generating software through conversational AI alongside risks involving reliability, security and limited user understanding.
What Is a Side-Channel Attack? How Computers Can Leak Secrets Without Being Hacked
Article link | Tech Xplore | Tech Xplore | July 30, 2026
Computer-science explainer describes how timing, power use, sound and other indirect signals can expose protected information without conventional system intrusion.
Claude AI Finds Cryptography Weaknesses Human Experts Missed
Article link | Paul Arnold | Tech Xplore | July 29, 2026
Anthropic researchers report that an experimental Claude model identified mathematical shortcuts and weaknesses in cryptographic systems that had eluded human specialists.
Stable-GFlowNet Uncovers More Hidden Weaknesses in Generative AI Models
Article link | Tech Xplore | Tech Xplore | July 29, 2026
Researchers developed a safety-testing method that finds substantially more hidden failure modes in generative AI systems than existing approaches.
To Defend Your Software, First Teach AI to Break It
Article link | Tonia Moxley | Tech Xplore | July 9, 2026
Cybersecurity researchers propose training artificial intelligence to identify and exploit software vulnerabilities so defenders can discover weaknesses before attackers do.
