NIH TO HALT FOREIGN SUBAWARDS BY SEPTEMBER 2025
The NIH announced it will halt “foreign subawards” by the end of September 2025, a policy change that affects U.S. researchers sharing grant funds with foreign collaborators. The NIH justifies the move as necessary for tracking funds and maintaining national security, though existing foreign funding will not be retroactively affected. The change is expected to disrupt collaborations on critical global health issues like tuberculosis, malaria, and HIV. Researchers argue the policy undermines scientific collaboration, with many calling it counterproductive. The decision follows scrutiny over a grant to EcoHealth Alliance linked to the Wuhan Institute of Virology. Critics, including former NIH officials, warn that halting foreign subawards will harm international research, increase costs, and hinder the sharing of knowledge. NIH’s new approach mandates direct applications from foreign scientists, which could complicate collaborations and create inefficiencies. The policy, scheduled to take effect by September 2025, is seen by many as harmful to global scientific progress. (Science)
AI BOT ERRORS HIGHLIGHT RELIABILITY ISSUES WITH PROGRAMMING TOOLS
An AI bot used by Cursor, a tool for programmers, mistakenly announced a non-existent policy change, causing customer outrage. The issue highlights a broader problem with AI’s reliability. Despite advancements in AI technology, particularly reasoning systems, errors known as “hallucinations” are becoming more frequent. These systems, such as those from OpenAI, Google, and DeepSeek, often produce inaccurate information due to their reliance on mathematical probabilities rather than definitive facts. The hallucination rates in recent AI models are as high as 79%, especially in complex tasks. While these bots excel in tasks like math and programming, they struggle with ensuring factual accuracy, causing problems in sensitive fields like law or healthcare. Despite efforts to improve accuracy, the causes of these hallucinations remain unclear, and the challenge persists. Researchers are working on ways to trace AI behavior, but the systems’ complex nature makes it difficult to fully understand their mistakes. (New York Times)
ACCELERATEX COFOUNDER PLANS AI-DRIVEN FEDERAL JOB AUTOMATION
Anthony Jancso, a cofounder of AccelerateX, a government tech startup, is hiring technologists for a project aimed at automating federal tasks using AI agents, potentially replacing tens of thousands of federal employees. Jancso, a former Palantir employee, revealed in a Slack post that his team plans to deploy AI agents across federal agencies, automating up to 70,000 jobs. The proposal has faced significant backlash, with critics expressing concerns about the feasibility and ethical implications of replacing human workers with AI. AccelerateX, which previously focused on solving social issues with AI, pivoted to government automation, securing a partnership with Palantir. Despite claims of job efficiency, experts argue that AI is unlikely to fully replace such a large workforce, citing challenges in implementing AI across diverse federal agencies and the technology’s unreliability. Jancso’s project is closely tied to the Department of Government Efficiency (DOGE), which seeks to use AI to streamline government operations. (Ars Technica)
GROK AI CHATBOT DISAPPOINTS CONSERVATIVE USERS WITH NEUTRALITY
Grok, the AI chatbot created by Elon Musk’s xAI, was designed to appeal to conservative users by being an “anti-woke” alternative to other chatbots. However, its refusal to spread election misinformation or endorse controversial views has disappointed many, particularly MAGA supporters. While Grok was trained to align with conservative ideals, it often provides fact-based, neutral responses, which contradict some of the Trump administration’s key claims. This has led to frustration among users who expected Grok to reaffirm their beliefs. In one viral exchange, Grok explained that as it becomes smarter, its answers aim for truth and nuance, which often clashes with MAGA expectations. Despite Musk’s efforts to shape Grok’s responses, including allowing it to use unfiltered language, the chatbot’s neutrality and fact-checking have continued to frustrate those seeking partisan content, highlighting the challenges of balancing truth with ideological preferences. (Gizmodo)
CYBERSECURITY CONCERNS RAISED OVER EASYJSON’S RUSSIAN CONNECTION
Cybersecurity researchers have raised concerns about easyjson, an open-source software package widely used across U.S. sectors, including the Department of Defense. The tool, maintained by Russian developers linked to VK Group—led by Vladimir Kiriyenko, a sanctioned executive with close ties to the Kremlin—could pose a national security threat. While no vulnerabilities have been identified in easyjson, its connection to a sanctioned company and Russia’s history of cyberattacks on U.S. infrastructure raises alarm. Easyjson, a key component in cloud-native systems, has been hosted on GitHub since 2016 and is crucial for software in finance, healthcare, and technology. The risk stems from the potential for Russia to manipulate the code for espionage or cyberattacks. Experts warn that, although the software itself is efficient, its geopolitical links make it a potential tool for future threats. The rise of such risks has prompted greater scrutiny of open-source software’s provenance and the impact of foreign involvement. (Wired)
U.S. MEASLES OUTBREAK REACHES 935 CASES ACROSS 30 STATES
The U.S. has reported a total of 935 measles cases across 30 states, with 51 new cases in the latest CDC update. The majority of these cases are linked to a large outbreak centered in West Texas, where 683 cases have been reported. The outbreak has led to 89 hospitalizations and 2 deaths. The CDC has warned against alternative treatments for measles, such as vitamin A, antibiotics, and inhaled steroids, that have been promoted by Health and Human Services Secretary Robert F. Kennedy Jr. Vitamin A can reduce mortality in areas with vitamin A deficiency, but its overuse has led to hospitalizations in the U.S., where deficiency is rare. The CDC also confirmed that antibiotics are ineffective for treating measles, and steroid use should be based on individual health assessments. Meanwhile, New Mexico reported an additional case, and health officials continue to urge vaccinations ahead of summer events. (CIDRAP)
LOUISIANA COASTAL RESTORATION PROJECT FACES NEW CHALLENGES AND DELAYS
The $3 billion Mid-Barataria Sediment Diversion project in Louisiana, aimed at restoring a rapidly eroding coastline, has faced new challenges due to accusations from Gov. Jeff Landry that his predecessor, John Bel Edwards, concealed a critical study. The study, which projected the diversion would create less land than other models, was not disclosed to the U.S. Army Corps of Engineers, potentially compromising the project’s federal approval. Edwards denied the claims, while Landry criticized the project, claiming it would harm Louisiana’s fishing culture. The controversy intensified when the Corps suspended the project’s permit, citing the withholding of information as a key factor. Despite opposition from local fishing groups, who fear the diversion’s impact on the industry, environmental advocates warn that halting or altering the project could undermine future infrastructure efforts and disrupt crucial restoration work funded by BP’s Deepwater Horizon settlement. The state now faces the possibility of losing federal funding if the project is further delayed or abandoned. (AP)
WISCONSIN MAN’S SNAKEBITE TOLERANCE STUDIED FOR ANTIVENOM DEVELOPMENT
Tim Friede, a Wisconsin man who has purposefully been bitten by snakes hundreds of times, is now the subject of scientific study aimed at creating better treatments for snakebites. Friede, who developed a fascination with venomous creatures, began injecting himself with small doses of snake venom to build up tolerance. Over nearly two decades, he withstood numerous bites, including from black mambas, taipans, and water cobras, hoping to push his body to its limits. His immune system has developed antibodies that may neutralize venom, a process researchers are now studying. Despite the dangerous nature of his methods, his blood contains antibodies that could be used in antivenom production. A recent study identified two antibodies that neutralize venom from various snakes, but the treatment is still in early stages, with trials in mice showing promise. While this research could potentially save lives, Friede warns others not to follow his dangerous example. (AP)
ASTROAGENTS AI SYSTEM DEVELOPED TO ADVANCE ASTROBIOLOGY RESEARCH
AI researchers have developed AstroAgents, an autonomous system designed to assist in astrobiology by generating scientific hypotheses and analyzing data. This tool comprises eight AI agents capable of studying Mars samples, particularly focusing on identifying organic molecules that may indicate past or present life. AstroAgents utilizes agentic AI, a more active form of AI, which is capable of formulating hypotheses, evaluating results, and adapting its methods. The system recently generated 101 hypotheses related to potential biomarkers for life, some of which were deemed novel and plausible. Experts believe AstroAgents can help analyze complex data sets, like mass-spectrometry plots, beyond human capability. However, some scientists, such as Michael Wong, question the tool’s true contribution, suggesting that it remains unclear whether these hypotheses are groundbreaking. Despite the debate, proponents like Denise Buckner view AstroAgents as a promising tool for studying Mars samples in future missions and advancing the understanding of life’s origins. (Nature)
Thanks for reading. Let’s be careful out there.
WORDS: The Biology Guy.
Leave a Reply