The U.S. government and AI companies have been working together to set rules for powerful AI models, but the process has been complicated and unclear. The Trump administration changed earlier safety rules from the Biden era and imposed export controls and licensing on companies like Anthropic, while some experts say better planning and more technical knowledge in government could have made regulation smoother.
Key Facts
OpenAI and Anthropic received government approval to release powerful AI models under President Trump’s regulatory approach.
The Trump administration removed some Biden-era rules that required companies to report AI safety test results to the government.
Export controls and licensing requirements were applied to Anthropic due to safety and security concerns, especially related to “jailbreaking” (tricking AI to bypass safety limits).
Experts say there was no clear framework to decide how serious AI safety issues must be before government action.
The U.S. government has struggled to hire AI experts; less than 1% of AI PhDs work for the government.
Agencies involved in AI regulation, like the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation, are underfunded.
Some states have started their own AI rules and training programs, but Congress has not passed comprehensive AI safety laws.
AI industry leaders, including OpenAI’s CEO Sam Altman, have found the government negotiation process productive despite challenges.
Read the Original
Want the full story? Tap a source to open the original
article.
Scientists at UCLA discovered that creatine, a supplement often used for muscle building, may help the immune system fight cancer. Their study found that creatine supports special immune cells called dendritic cells, which activate cancer-fighting T cells.
Key Facts
Creatine is commonly used by people who exercise to build muscle.
Researchers studied mice and human cells to understand creatine’s effect on the immune system.
Dendritic cells use creatine to work properly inside tumors.
These cells act like “intelligence officers” that help the body recognize and attack cancer.
When dendritic cells could not absorb creatine, they were weaker and less able to activate T cells.
T cells are immune cells that can kill cancer cells once activated.
The study suggests that creatine might improve the body’s response to cancer immunotherapy.
More research, including clinical trials, is needed to confirm these results in humans.
Read the Original
Want the full story? Tap a source to open the original
article.
US Senator Ed Markey has introduced a set of bills called the "AI accountability agenda" to address the dangers of unregulated artificial intelligence (AI). His proposed laws focus on controlling the environmental impact of AI data centers, preventing workplace discrimination by AI, protecting children from harmful chatbots, and ensuring human oversight of AI decisions.
Key Facts
Senator Ed Markey plans new federal rules requiring AI data centers to get approval from the Federal Communications Commission (FCC) before construction.
The FCC would check that data centers do not harm air and water quality, local wildlife, noise levels, energy costs, and jobs.
Markey aims to stop companies from mainly using AI systems to hire, fire, or promote workers.
There would be stronger rules to protect kids from emotional dependence on AI chatbots.
AI developers would need to do independent checks to find bias or discrimination before releasing AI tools that affect important decisions.
Federal agencies using AI must have offices focused on fighting bias and discrimination.
Health care facilities would be required to have options for humans to override AI decisions.
Workers would be protected if they disagree with AI recommendations in their jobs.
Read the Original
Want the full story? Tap a source to open the original
article.
Artificial intelligence (AI) is creating a big gap in how people experience and use it. While some experts use advanced AI tools for coding and research, most Americans only see simpler versions like chatbots or search helpers and don’t fully understand or trust AI.
Key Facts
Advanced AI models like OpenAI’s Sol and Anthropic’s Fable are used by a small group of expert users for tasks like coding and cybersecurity.
Most Americans encounter AI through basic tools like chatbots, search summaries, or customer service bots, not advanced AI systems.
Around half of U.S. adults use AI chatbots, mostly for simple information searches.
There is a hierarchy of AI users: free users, paid users, power users, preview users, and an insider class with early access to new AI tools.
AI trust is low: 63% of Americans think AI is advancing too fast, and only 16% believe it will benefit society in the next 20 years.
The biggest benefits of AI currently go to investors, big tech companies, and elite users, while many workers face job disruptions.
The U.S. government has created AI literacy programs and funded workforce retraining to help workers adapt to AI changes.
Despite education efforts, most people lack access to the best AI tools and deeper technical knowledge.
Read the Original
Want the full story? Tap a source to open the original
article.
China successfully landed and recovered a reusable rocket booster for the first time, a major step forward in its space program. The Long March 10B rocket booster landed vertically on a floating platform about six minutes after launch from Hainan, southern China.
Key Facts
The Long March 10B rocket launched from Hainan, China, at 12:15 local time.
About six minutes after separating from the upper stage, the booster landed vertically on a floating platform.
This is China’s first successful landing of a reusable rocket booster.
Reusable boosters help reduce the cost of space launches by allowing important parts to be used multiple times.
The Long March 10B can carry a payload of at least 16 metric tons to low-Earth orbit.
Instead of landing on a ground pad or drone ship like SpaceX’s Falcon 9, the Long March 10B uses landing hooks to catch a net on a floating platform.
China’s earlier attempt in February 2024 with the Long March 10A resulted in a controlled splashdown near a recovery platform.
Shares in Chinese space companies rose sharply after the news, hitting the daily allowed limit increase.
Read the Original
Want the full story? Tap a source to open the original
article.
Telstra experienced a nationwide outage that lasted nearly five hours due to a software fault in its time-keeping system, which caused the network to think it was 2006. The company apologized and is investigating why backup systems failed to prevent the problem. Meanwhile, South Australian police are looking into whether a death during the outage was linked to emergency call failures.
Key Facts
Telstra's network went down for almost five hours on Wednesday morning because its time system reset to the wrong date (2006).
The outage affected mobile phone service, payment machines (Eftpos), rail services, and other systems depending on the network.
Telstra knew about risks related to its time-keeping system before the outage.
Backup systems designed to stop such failures did not work, and Telstra is investigating why.
Telstra CEO Vicki Brady apologized and spoke publicly after cutting short her holiday.
Police in South Australia are investigating if a person's death during the outage was linked to emergency call problems.
A similar time-related outage happened to another telecom company in Jersey in 2020, which also disrupted emergency calls.
The error occurs when a system's internal clock resets to a date far in the past, causing network equipment to stop working properly.
Read the Original
Want the full story? Tap a source to open the original
article.
Online ads in the UK are promoting small, portable air conditioners that claim to cool rooms quickly and efficiently. The Advertising Standards Agency says these claims are often false, and testing showed the products are usually ineffective, sometimes just simple fans.
Key Facts
Ads appear on Facebook, YouTube, and other platforms, claiming these portable air conditioners cool rooms in 90 seconds.
The Advertising Standards Agency (ASA) warns many ads make unrealistic claims and use fake customer reviews.
Devices tested by a YouTuber cost about £70 but worked no better than a cheap fan.
Some ads use AI-generated images to make products look more advanced.
These devices are often “swamp coolers” that work poorly in humid climates like the UK’s.
The ASA advises consumers to be cautious of products that promise fast cooling, have poor grammar in ads, or feature too-good-to-be-true reviews.
People can report suspicious ads to the ASA for investigation.
The ASA monitors these ads and can ban them, but it cannot fine advertisers directly.
Read the Original
Want the full story? Tap a source to open the original
article.
A Singapore-based company, Datagrid, plans to build New Zealand’s first AI datacentre near Invercargill, with construction starting this year and operations by 2028. Local residents are concerned about the project's environmental impacts, including high electricity and water use, noise, and limited long-term local benefits, while authorities highlight potential job creation and economic growth.
Key Facts
Datagrid received approval to build a NZ$3.5 billion AI datacentre on a 49-hectare site in Makarewa, near Invercargill.
The datacentre will serve AI and cloud providers globally and plans to build a high-speed internet cable to Australia.
It will be New Zealand’s second largest electricity user, consuming 280 megawatts.
Construction will create about 1,200 jobs, with around 50 permanent positions after completion.
Residents worry about environmental effects like high water use, noise pollution, and the use of up to 84 diesel generators during power shortages.
Some experts doubt the long-term economic benefits for the local area, pointing out limited permanent jobs.
The New Zealand government supports datacentre projects to attract NZ$25-30 billion in foreign investment and grow AI infrastructure.
A report estimates the datacentre industry could generate up to NZ$70 billion in economic activity in the next decade.
Read the Original
Want the full story? Tap a source to open the original
article.
Fidji Simo is leaving her role as OpenAI’s head of product and business due to worsening health from a chronic illness called POTS. She will continue to advise OpenAI part-time while the company’s president is handling product responsibilities during her absence.
Key Facts
Fidji Simo stepped down because of a serious worsening of Postural Orthostatic Tachycardia Syndrome (POTS), a condition diagnosed in 2019.
She began medical leave in April before officially resigning from her full-time role.
OpenAI President Greg Brockman took over product duties during her medical leave.
Simo will stay at OpenAI as a part-time advisor after stepping down.
CEO Sam Altman expressed sadness about her departure and wished her a quick recovery.
Simo joined OpenAI in 2025 after leading Instacart through its public offering and working over ten years at Meta.
She was hired to help grow OpenAI’s business as the company moved from research to real-world AI products.
The news was first reported by The Wall Street Journal.
Read the Original
Want the full story? Tap a source to open the original
article.
Meta has introduced Muse Image, an AI tool that can create pictures using faces from public Instagram profiles by default. Privacy experts are concerned because users are not notified when their photos are used, and opting out requires finding a hidden setting.
Key Facts
Muse Image allows AI-generated photos using faces from public Instagram accounts.
Private accounts and users under 18 are automatically excluded from the feature.
Adult users with public accounts can opt out by changing a setting in Instagram's “sharing and reuse” section.
Meta does not notify users when their photos are used in AI-generated images by default.
Privacy experts say the opt-out setting is hard to find and that data sharing is turned on by default.
Meta says it has safety controls and will act against content violating community rules.
There is uncertainty if children shown in public photos of adults are included in AI image generation.
Critics argue this setting should be opt-in since users did not expect new uses of their photos when joining Instagram.
Read the Original
Want the full story? Tap a source to open the original
article.
Several news outlets, including The New York Times, have asked a judge to punish OpenAI for not providing evidence in a legal dispute. The case involves claims that OpenAI used copyrighted news content without permission to train its AI models.
Key Facts
Multiple news organizations are involved in the lawsuit against OpenAI.
They claim OpenAI used their copyrighted articles without approval.
The issue centers on how OpenAI trained ChatGPT's AI technology.
The news outlets say OpenAI did not fully share evidence in court.
They want the judge to impose sanctions on OpenAI for withholding information.
The dispute highlights conflicts over AI use of copyrighted material.
OpenAI denies it can search some data to respond to the legal requests.
Read the Original
Want the full story? Tap a source to open the original
article.
Meta has introduced a new AI tool called Muse that lets Instagram users create AI-generated images using photos from public accounts. This tool can make people’s images vulnerable to deepfakes without their permission or awareness.
Key Facts
Meta launched an AI imaging tool named Muse.
Muse can create AI-generated images using photos from public Instagram profiles.
Users do not need to get approval or notify the person whose photo is used.
This raises concerns about people being targeted by deepfakes—fake images or videos made to look real.
The tool could affect the privacy and security of public Instagram users.
Muse is part of a growing trend in AI tools that manipulate images.
Read the Original
Want the full story? Tap a source to open the original
article.
OpenAI has launched ChatGPT Work, a new tool designed to help users complete long and complex projects by automating workflows and managing tasks over time. The tool connects with common work apps, can run scheduled tasks, and uses updated AI models to improve performance while allowing users to approve important actions.
Key Facts
ChatGPT Work can handle tasks for hours, turning goals into finished work.
It integrates with workplace tools like Slack, Microsoft Teams, Google Drive, and SharePoint through plugins.
The tool includes Scheduled Tasks, which automate repetitive jobs on a set schedule or when certain events happen.
Users can control access to outside apps and files with compliance and admin settings.
ChatGPT Work merges with OpenAI’s Codex coding app technology but offers separate views for different uses.
The existing basic ChatGPT app is renamed “ChatGPT Classic” and is now less prominent.
Usage of ChatGPT Work is billed based on a credit system and subscription plans, with usage limits to control costs.
The launch coincides with the release of the new GPT-5.6 AI model, designed for better cost efficiency and stronger performance.
Read the Original
Want the full story? Tap a source to open the original
article.
Several U.S. media companies, including The New York Times and Daily News, have asked a federal court to penalize OpenAI for allegedly hiding evidence in a lawsuit about copyright. The case focuses on whether OpenAI used millions of news articles without permission to train its AI, ChatGPT, which may affect how news is shared and paid for online.
Key Facts
The New York Times, Daily News, and other media outlets are suing OpenAI and Microsoft over copyright issues related to AI.
The media companies claim OpenAI hid important data showing how it used news articles to train ChatGPT.
They want the court to sanction OpenAI for "discovery misconduct," which means hiding or destroying evidence.
OpenAI says revealing ChatGPT conversation logs could harm user privacy.
The lawsuit started in late 2023 after ChatGPT’s rise changed how people find information on the internet.
Google introduced AI summaries in 2024, reducing traffic and ad money for news websites.
Other copyright owners like authors and artists are also suing AI companies for using their work without permission.
The New York Times has spent over $28 million on court cases against AI companies, while some media companies have licensed their content to AI firms.
Read the Original
Want the full story? Tap a source to open the original
article.
Microsoft released a patch on Wednesday to fix a security flaw in Windows Defender that allowed attackers to take control of Windows 10 and 11 systems. However, this patch may cause a new problem where the system can fill up the hard drive by writing very large files, leading to disk space running out and system instability.
Key Facts
The patched flaw, called RoguePlanet (CVE-2026-50656), lets attackers control Windows machines remotely.
Microsoft’s update fixes this vulnerability automatically through the Microsoft Malware Protection Engine.
The new patch includes extra security updates called “defense-in-depth.”
A security researcher named NightmareEclipse says these updates cause a bug that can make Defender write huge files that fill up the disk.
This happens because Defender tries to keep copies of special hidden metadata files called Zone.Identifier for files downloaded from the internet.
Attackers could exploit this through a custom SMB server, which is a way computers share files over a local Windows network.
The bug causes Defender to lock files and fill the disk space, making the computer unstable but not crash.
Microsoft has not confirmed whether the disk-filling problem is real yet.
The researcher and Microsoft have had ongoing disagreements about vulnerability disclosures and patches since May.
Read the Original
Want the full story? Tap a source to open the original
article.
Surgeons remotely controlled humanoid robots to remove gallbladders from live pigs in a new medical experiment. This trial shows potential for cheaper, smaller surgical robots that could be used in places with fewer resources.
Key Facts
The experiment involved removing gallbladders from live pigs using humanoid robots controlled by surgeons.
The study was published in the journal Nature and conducted by researchers at UC San Diego.
The robots used were Unitree G1 humanoid robots, made by a Chinese company, costing much less than current specialized surgical robots.
Unitree G1 robots weigh about 60 pounds and stand 5 feet tall, much smaller than existing surgical robots that weigh around 1,800 pounds.
Surgeons operated the robots remotely using a computer console with a headset and foot pedal.
The system still requires improvements because surgery took longer and needed pauses to recalibrate the robots.
The robots’ smaller size limits their movement and reach compared to human arms.
If developed further, this technology could help hospitals in rural areas, battlefields, or even space missions.
Read the Original
Want the full story? Tap a source to open the original
article.
A writer named Rose Horowitch says that fewer Americans are reading books now. She believes that new technology and shorter attention spans have made gambling and other activities more popular than reading.
Key Facts
Rose Horowitch wrote an article for The Atlantic in August.
She claims that the era when many people read books is ending.
Gambling has become more popular as a leisure activity than reading.
New technology is partly responsible for shorter attention spans.
Overall, fewer Americans are reading and literacy levels are dropping.
Read the Original
Want the full story? Tap a source to open the original
article.
OpenAI is involved in a legal dispute with news organizations, including The New York Times, over the use of news content by its AI chatbot, ChatGPT. The news groups accuse OpenAI of hiding evidence related to users bypassing paywalls, which has led to calls for penalties against OpenAI.
Key Facts
News organizations, led by The New York Times, are suing OpenAI over copyright issues with ChatGPT using their news articles.
OpenAI reportedly hid information and made false statements about its ability to search ChatGPT’s user logs.
These logs show whether users accessed news content behind paywalls through ChatGPT.
OpenAI had two large samples of anonymized logs but did not disclose them for two years.
The company had already searched these logs to develop a filter to block copyrighted content.
News plaintiffs say OpenAI’s actions delayed the legal process and increased costs.
OpenAI denies wrongdoing and says the lawsuit is an attempt to invade user privacy.
The dispute centers on whether ChatGPT’s use of news content is fair use or infringement.
Read the Original
Want the full story? Tap a source to open the original
article.
OpenAI released its new AI model, ChatGPT 5.6, after delaying the launch due to US government cybersecurity concerns. The Trump administration asked OpenAI to limit access to trusted users first and review the model before a wider release.
Key Facts
OpenAI delayed ChatGPT 5.6 to address cybersecurity issues raised by the Trump administration.
The administration requested a limited release to government-approved users initially.
The government’s Center for AI Standards and Innovation tested the model before wider access was allowed.
Anthropic, a rival AI company, also faced US restrictions and an export ban on its AI models last month.
The Trump administration supports advancing AI to compete with China but has recently increased security measures due to cyber-attack risks.
President Trump issued an executive order encouraging voluntary reviews of AI models by the government.
US restrictions on advanced AI models have caused concern among businesses, foreign governments, and US allies.
OpenAI and Anthropic are competing ahead of planned stock market offerings, with OpenAI targeting a $1 trillion valuation.
Read the Original
Want the full story? Tap a source to open the original
article.
A new report shows strong growth in local groups opposing large AI data centers across the U.S., with over 430 groups and more than 525,000 members in 40 states. People are concerned about the centers’ energy use, water demand, noise, and other impacts as these facilities expand rapidly.
Key Facts
The number of opposition groups rose from 268 in April to 430 in the latest report.
Opposition members increased nearly seven times since December 2025, now totaling over 525,000 people.
Ohio has the most groups (47) while Pennsylvania has the second most (46), but Missouri has the largest number of members (about 141,000).
The report tracks groups mainly on Facebook to avoid counting duplicates on other platforms.
Data centers use lots of energy, sometimes as much as major cities, driven by demand for AI technology.
There are nearly 4,500 data centers in the U.S., including almost 500 in Texas and over 637 in Virginia.
A national protest against data centers is planned for July 18 in over 50 locations.
Concerns against data centers include rising utility costs, high water use, noise, land changes, and national security risks related to AI growth.
Read the Original
Want the full story? Tap a source to open the original
article.
