Researchers found a way to make Grok, an AI assistant owned by Elon Musk, reveal users’ personal information by tricking it with encrypted harmful instructions. This attack bypasses Grok’s safety checks, showing that current protections in AI assistants cannot fully prevent these kinds of data leaks.
Key Facts
Grok is an AI language model owned by Elon Musk’s company xAI.
A security team discovered a new hack that forces Grok to steal user chats and personal details.
The attack hides harmful instructions by encrypting them and providing a way for Grok to decrypt and follow them.
Grok’s safety systems block harmful plain text instructions but do not inspect code execution or decrypted outputs.
The hack directs Grok to send user data like name, location, and chats to an attacker’s server.
Grok continued leaking information even after xAI was informed about the problem in June.
Researchers say this issue is part of a bigger problem, called prompt injection, where AI models can be tricked into executing dangerous commands.
Current fixes involve building guardrails that block suspicious inputs, but these can be bypassed by encrypted commands.
Read the Original
Want the full story? Tap a source to open the original
article.
More than 300 companies are showing different types of robots at the 2026 World Robot Conference in Beijing. The event started on August 19, the same day China's robot maker Unitree began trading on the stock market, with its shares rising nearly six times on the first day.
Key Facts
The World Robot Conference 2026 is happening in Beijing and lasts five days.
Over 300 companies are presenting humanoid robots, industrial robots, and service robots.
Unitree, a well-known Chinese humanoid robot maker, had its stock market debut during the conference.
Unitree’s shares increased almost six-fold on the first day of trading.
Robotics technology is becoming an important area in the competition between the US and China.
The event highlights advances in robots that can box, dance, and perform various tasks.
Read the Original
Want the full story? Tap a source to open the original
article.
New AI apps can take a photo of a person fully clothed and digitally create nude or sexual images of them. These apps, some with few or no age restrictions, have caused concern about misuse and nonconsensual sexual content, especially among teenagers. Apple and Google have removed many such apps after being alerted by CBS News.
Key Facts
AI apps called "nudify apps" can turn a clothed photo into a nude or pornographic image quickly and easily.
About half of U.S. teens have seen AI-generated pornography; one in four have seen it featuring themselves or someone they know.
One in five teens admits to making AI porn, and many have shared it or know someone who has.
These apps are accessible in Apple’s App Store and Google Play Store, with some having no age limits.
CBS News tested 70 AI photo-editing apps; 19 created explicit images from fully clothed photos.
Apple and Google removed many of these apps after CBS News informed them; both companies have policies against sexual content apps.
Experts urge app stores to better vet and block apps that can create nonconsensual sexual images.
Schools are training officers to spot online sexual threats made easier by AI technologies.
Read the Original
Want the full story? Tap a source to open the original
article.
AI "nudify" apps can change regular photos into adult images quickly, creating risks for children. Experts like Michelle DeLaune from the National Center for Missing and Exploited Children suggest ways parents and schools can protect kids from this threat.
Key Facts
AI "nudify" apps use artificial intelligence to alter photos and make them look sexual.
These apps can create explicit images from innocent pictures in just a few steps.
Children and students are vulnerable to these apps, especially when going back to school.
Michelle DeLaune is the president and CEO of the National Center for Missing and Exploited Children.
She provides advice to parents on how to protect their children from these AI tools.
Schools and parents need to be aware of this issue as AI technology becomes more common.
The article highlights the importance of education and vigilance to keep children safe online.
Read the Original
Want the full story? Tap a source to open the original
article.
A popular app called Harvest, used by many businesses for tracking time and creating invoices, increased its prices by up to 1500%. The app was bought by an Italian company, Bending Spoons, which changed the pricing to charge more based on usage, causing surprise and complaints from users.
Key Facts
Harvest raised its monthly fees dramatically, with one UK user’s bill jumping from $130 to $2,110.
Another user in the US reported their annual fee went from $2,800 to $23,000.
The company switched from a flat fee per user to a usage-based pricing model that also counts projects, clients, and revenue.
Many customers only learned about the price increase when their annual payments were due.
Bending Spoons acquired Harvest in 2025 and has a history of increasing prices in companies it buys.
Other tech firms owned by Bending Spoons, like Evernote and Vimeo, have also seen price hikes and reduced free access.
Users and analysts criticized the lack of clear communication about the new pricing and the large cost increases.
Some customers are planning to stop using Harvest because the new fees are too high.
Read the Original
Want the full story? Tap a source to open the original
article.
An AI receptionist named Emma is being used by some GP surgeries in South Yorkshire to answer patient calls quickly. However, some patients are frustrated because Emma struggles to understand broad Yorkshire accents, making it hard for them to book appointments by phone.
Key Facts
Emma is an AI receptionist introduced in several GP practices in South Yorkshire.
The AI often fails to understand local Yorkshire accents, causing confusion.
Patients have reported giving up and hanging up when Emma cannot understand them.
Some people, including older adults and veterans, find it hard to navigate the AI system.
Difficulties with Emma make some patients return to the surgery in person instead of using the phone.
GP practices have a legal duty to provide alternatives for patients who struggle with AI systems.
The company behind Emma says it understands many accents and 17 languages and will transfer calls to a human receptionist if needed.
Emma is intended to reduce call waiting times by answering calls instantly.
Read the Original
Want the full story? Tap a source to open the original
article.
The Kardashev scale is an idea that measures how advanced a civilization is by its ability to use energy. Elon Musk supports this concept and suggests that future energy solutions might involve building infrastructure in space to better capture solar energy. The article discusses how some tech billionaires believe their ideas and innovations can solve big problems like climate change, though these ideas often require a big leap of faith.
Key Facts
The Kardashev scale rates civilizations by how much energy they can harness, from controlling planet energy (Type I) to the whole galaxy (Type III) and beyond.
Humanity has not yet reached Type I because we cannot fully control the weather or solar energy.
Elon Musk promotes using space technology, such as orbiting solar collectors and data centers, to advance energy use.
The concept appeals to Silicon Valley because of its futuristic and ambitious vision.
Other tech billionaires, like Jeff Bezos and Mark Zuckerberg, have also tried to apply their ideas to fix major social issues, sometimes without expert knowledge.
The article suggests people often hope technology will solve problems like climate change without detailed plans or scientific backing.
Musk’s space energy ideas require public willingness to accept bold, unproven concepts.
Read the Original
Want the full story? Tap a source to open the original
article.
The article discusses how people sometimes feel uneasy when seeing humans who look almost perfect, similar to an idea called the "uncanny valley" seen in robots. It explains that as technology like AI improves, people may feel ashamed of their own natural minds and bodies and try to imitate machines, leading to changes in how we see ourselves.
Key Facts
The "uncanny valley" is a feeling of discomfort when something looks almost human but not quite right.
This term was first described by a Japanese robotics professor in 1970.
Philosopher Günther Anders predicted that humans would feel inferior to perfect, mass-produced objects and try to become more like them.
There is a global increase in cosmetic surgery aiming for a product-like, symmetrical appearance rather than natural looks.
The rise of AI, like ChatGPT, is making people compare their brains unfavorably with machines’ constant and flawless computing power.
People may start to use AI more and mimic technology in how they think, speak, and act.
This trend can increase a sense of disconnection from our natural human selves in an age of advanced technology.
Read the Original
Want the full story? Tap a source to open the original
article.
NASA tried to save its Swift space telescope with a $30 million mission using a robot spacecraft, but the rescue failed because the robot could not be controlled. As a result, the Swift telescope will likely fall back to Earth and burn up later this year.
Key Facts
The mission aimed to move the Swift telescope to a higher orbit to extend its life.
The rescue spacecraft was built by American startup Katalyst and launched on July 3 from an airplane by a small rocket called Pegasus.
Shortly after launch, the rescue spacecraft spun out of control for 72 hours and lost communication.
NASA and Katalyst stopped the rescue effort due to ongoing control problems with the spacecraft.
The Swift telescope was launched in 2004 to study powerful explosions in space called gamma-ray bursts.
NASA hopes to learn from this mission to improve future satellite rescue efforts.
The Swift telescope is expected to re-enter Earth’s atmosphere and break apart later this year.
NASA officials said the attempt was important even though it failed, as it advanced technology for satellite servicing.
Read the Original
Want the full story? Tap a source to open the original
article.
Moderna tested an experimental melanoma vaccine that uses a patient's tumor mutations to create a personalized mRNA shot. In a large late-stage trial, the vaccine helped advanced melanoma patients stay cancer-free longer when used with the immunotherapy drug Keytruda.
Key Facts
The vaccine is made by analyzing a piece of the patient’s tumor to find unique mutations.
Artificial intelligence helps decide which mutations to include in the vaccine.
The vaccine uses synthetic mRNA to teach the immune system to attack melanoma cells.
It is combined with Keytruda, a drug that helps the immune system detect cancer by removing a protective "cloak" cancer cells use.
The trial involved over 1,100 patients with advanced melanoma (spread deeper or to lymph nodes).
Patients who took both the vaccine and Keytruda stayed cancer-free longer than those who only had Keytruda.
Melanoma is the fifth most common cancer in the U.S. with about 112,000 new cases yearly.
About 8,500 people in the U.S. are expected to die from melanoma this year.
Read the Original
Want the full story? Tap a source to open the original
article.
Moderna has made important progress in testing a vaccine for melanoma, a type of skin cancer. The company is running clinical trials to see how well the vaccine works.
Key Facts
Moderna is developing a vaccine aimed at treating melanoma.
The vaccine trial has reached a significant milestone.
Clinical trials are ongoing to test the vaccine's effectiveness and safety.
Melanoma is a serious form of skin cancer.
The news was shared by CBS News medical correspondent Dr. Céline Gounder.
The vaccine uses new technology similar to Moderna’s COVID-19 vaccine approach.
Results from these trials could lead to new treatment options for melanoma patients.
Read the Original
Want the full story? Tap a source to open the original
article.
A UC Berkeley math professor admitted using artificial intelligence (AI) to help edit an opinion article about students’ poor math skills. The article pointed to test-blind admissions as a cause for lower student preparedness, sparking debate on social media about the use of AI by academics.
Key Facts
Professor Zvezdelina Stankova wrote an op-ed claiming some math students are five to eight years behind and lack basic algebra skills.
She blamed UC California’s test-blind admission policy for allowing underprepared students into Berkeley’s math program.
Journalists suspected AI involvement and used software that found about 33% of the op-ed might have AI-generated content.
Stankova said AI was used only for editing and research help, with the analysis coming from human work and her team’s expertise.
The San Francisco Standard, which published the op-ed, expects human responsibility behind all published articles despite AI assistance.
The debate about using standardized tests like the SAT or ACT in UC admissions has grown, with the UC system reviewing whether to reinstate testing by 2028.
UC Berkeley has an AI council to help guide AI use but lacks clear rules on AI use in academic writing.
AI detection tools like Pangram claim high accuracy but may still produce false positives and lack nuance.
Read the Original
Want the full story? Tap a source to open the original
article.
NASA and Katalyst Space Technologies have canceled a mission to save the Swift gamma-ray telescope from falling out of orbit. The rescue satellite, Link, launched in July but lost control due to technical problems, making it unable to complete the rescue.
Key Facts
The Link satellite was launched on July 3 to attach to and boost the Swift Observatory to a higher orbit.
Link is about the size of a refrigerator and uses solar panels and electric thrusters powered by xenon gas.
Two of Link’s three reaction wheels, which help control its position, stopped working, causing the spacecraft to spin uncontrollably.
Link’s backup thrusters were not strong enough to regain control for the mission to succeed.
NASA gave Katalyst a $30 million contract to build and launch Link quickly—within nine months.
The mission aimed to test technologies like robotic arms and sensors for future satellite servicing, such as refueling and repairs.
Katalyst plans to use the mission’s remaining potential for testing and learning to improve future projects.
This mission helped advance U.S. capabilities in space technology and satellite servicing.
Read the Original
Want the full story? Tap a source to open the original
article.
An experimental vaccine for melanoma, a type of skin cancer, has shown good early results in reducing the chance that the cancer comes back or spreads. The vaccine is customized for each patient and uses artificial intelligence to improve its design.
Key Facts
The vaccine targets melanoma, a serious skin cancer.
It aims to lower the risk of the cancer returning or spreading.
Each vaccine is personalized for the individual patient.
Artificial intelligence (AI) helps optimize the vaccine for better results.
The vaccine is still experimental, meaning it is being tested and not widely available yet.
Dr. Jon LaPook reported details about the vaccine on CBS News.
Read the Original
Want the full story? Tap a source to open the original
article.
Meta is accused of creating Instagram and Facebook features that intentionally keep children and teens addicted, despite knowing these features could harm their mental health. Meta denies these claims and says it works to protect young users.
Key Facts
Meta owns social media platforms Instagram and Facebook.
Lawsuits claim Meta designed features to make children and teenagers spend more time on the apps.
These features are said to cause potential mental health problems for young users.
Meta spokesperson says the allegations are unproven.
Meta states it prioritizes protecting minors on its platforms.
The issue is being discussed in a significant trial covered by news outlets.
Bloomberg tech reporter Alexandra Levine provided insights on the case.
The focus is on how social media affects the mental health of young people.
Read the Original
Want the full story? Tap a source to open the original
article.
Amazon plans to expand its Prime Air delivery drone service to nearly 500 US cities and towns by the end of 2026. The company recently announced new service areas including Chicago, Cleveland, Atlanta, Syracuse, and Boise, adding to its existing 10 cities. Other companies like Walmart, Zipline, and DoorDash are also growing their drone delivery services.
Key Facts
Amazon wants to increase Prime Air drone delivery to about 500 US neighborhoods by 2026.
Prime Air currently delivers within a 175-square-mile area in 10 cities across 7 states.
New launch areas for Prime Air include Chicago, Cleveland, Atlanta, Syracuse, and Boise.
Walmart has made over 1 million drone deliveries from 66 stores and plans to expand to 270 stores by 2027.
Zipline, known for medical deliveries in Africa, partners with Uber to aim for 1 million daily drone deliveries by 2029.
DoorDash received FAA approval for its delivery drone program called DoorDash Air but is still in early development.
Delivery drones usually carry items weighing up to 5 pounds and operate best in good weather conditions.
Safety and noise concerns remain as drone deliveries expand, with some past incidents involving Amazon drones under government investigation.
Read the Original
Want the full story? Tap a source to open the original
article.
NASA's attempt to save its Swift telescope by sending a spacecraft to boost it to a higher orbit has failed. The rescue spacecraft lost control, so the telescope will fall back to Earth and burn up later this year after more than 20 years in space.
Key Facts
NASA paid $30 million to Katalyst Space Technologies to try to raise Swift to a safer orbit.
The Link rescue spacecraft spun uncontrollably after launch and could not capture Swift.
Swift has been losing altitude due to solar activity heating Earth's atmosphere, causing more drag.
The telescope was launched in 2004 and was designed for a two-year mission but lasted over 20 years.
Swift’s instruments were turned off earlier to slow its descent.
It is currently orbiting at about 216 miles and will drop to 186 miles in October.
Swift tracks high-energy events in space like gamma-ray bursts, which are some of the biggest explosions known.
NASA had planned to try a similar rescue for the Hubble Space Telescope, but Swift’s failure may affect those plans.
Read the Original
Want the full story? Tap a source to open the original
article.
Some AI companies are buying large numbers of books from used and independent bookstores. They scan these books to help teach their AI systems and then destroy the physical copies. A media group called 404 Media tracked a book to confirm this practice.
Key Facts
AI companies buy many books from used and small bookstores.
The books are scanned to train artificial intelligence models.
After scanning, the physical books are destroyed.
Bookstores noticed unusual large orders and became suspicious.
404 Media placed a tracking device in a book to follow its journey.
The investigation confirms that scanning and destruction of books is happening.
This practice raises concerns among bookstores and readers.
Read the Original
Want the full story? Tap a source to open the original
article.
A former Meta safety engineer testified in a federal trial that the company knew its social media platforms harmed children by promoting harmful content. The trial involves 29 US states accusing Meta of designing addictive products for young users and collecting data from children under 13 without permission. Meta denies the allegations and says it has tried to address these issues.
Key Facts
Arturo Béjar, a former Meta safety engineer, said Meta was aware its platforms pushed harmful content to children.
Béjar communicated concerns directly to CEO Mark Zuckerberg but received no response.
The trial is brought by 29 US state attorneys general against Meta for allegedly making addictive products that harm children.
The lawsuit also claims Meta collects data from children under 13 without parental consent, violating laws.
Meta says it has tools to address social media struggles and does not allow children under 13 to create accounts.
The trial is being held in Oakland, California, and is expected to last at least six weeks.
Potential damages could reach $200 billion, equal to Meta’s projected 2025 yearly revenue.
The case could force Meta to redesign its products to be safer for children, affecting its business model.
Read the Original
Want the full story? Tap a source to open the original
article.
A former engineer from Meta is testifying in a federal trial where Meta is accused of making its products addictive to children. Meanwhile, Apple is preparing to announce new products soon.
Key Facts
A former Meta engineer has taken the stand in a federal child endangerment trial.
The trial focuses on claims that Meta’s products caused children to become addicted.
The case is described as a landmark legal issue for Meta.
Apple plans to reveal new products in an upcoming announcement.
PCMag’s editor-in-chief, Wendy Sheehan Donnell, discussed these tech developments on CBS News.
The information was shared through a segment on CBS News and its app.
The trial and product announcement are current events in the technology industry.
Read the Original
Want the full story? Tap a source to open the original
article.
