Category: Technology News

TAMPA, Fla. (AP) — The Trump administration is pushing to unleash the power of artificial intelligence for the U.S. military while facing calls to put up guardrails around the rapidly developing technology from some companies — and even notes of caution from top leaders in uniform.

Adm. Frank Bradley, head of U.S. Special Operations Command, told attendees of a recent annual special forces conference in Tampa, Florida, that troops “have to be very careful about how we come to (AI’s) employment and its inspiration into the delivery of lethality.”

Bradley said he can see a future where AI determines what targets to hit but that “we, as humans, have to have the confidence that … it’s going to deliver violence only where we intend it to be delivered.”

The remarks from Bradley, who oversees the units that handle the military’s most difficult and dangerous operations, about the need to ensure safeguards come as his boss, Defense Secretary Pete Hegseth, is pushing to rapidly evolve the military through AI. It is a push that has led to clashes with some tech companies worried about safety measures.

Hegseth has insisted that the Pentagon be allowed to use the technology any legal way it sees fit. He told an audience of SpaceX employees in January he would reject any AI models “that won’t allow you to fight wars” and that his vision for the technology was systems that operate “without ideological constraints that limit lawful military applications.”

AI’s use in the military is part of the Republican administration’s larger push to grow the capability it sees as a unique American advantage even as it faces pressure to ensure responsible safeguards.

President Donald Trump abruptly called off plans to sign a new AI executive order hours before an expected White House ceremony over concerns the measure could dull America’s edge on AI technology.

“We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead,” Trump told reporters.

Two differing AI worlds within the military

When asked about Bradley’s remarks, a Pentagon official said efforts are focused on using AI to create “functional battlefield tools” that can help troops come up with and identify targets more quickly and, as a result, speed up strikes on those targets. The official spoke on condition of anonymity to offer more candid remarks.

Officials at U.S. Special Operations Command talked about AI not as something that will help eliminate targets but rather as a tool that can offer troops more time to focus on their mission.

Sgt. Maj. Andrew Krogman, the top enlisted official for U.S. Special Operations Command, said at the conference that he sees AI handling administrative tasks to free up operators or helping modernize how the command does business.

Melissa Johnson, the top acquisition official for the command, said AI should be “reducing the cognitive workload on mundane tasks.”

“We’re leveraging AI more and more, but it’s not to replace operator judgment, it’s to enhance it,” she added.

Helen Toner, interim executive director at Georgetown University’s Center for Security and Emerging Technology, said those differing descriptions about AI in the military are both true.

“There are a huge number of potential uses for AI in these kinds of bureaucratic settings, which the U.S. military is actively exploring,” Toner said.

Lt. Gen. Michael Conley, head of Air Force Special Operations Command, told a congressional committee in May that his troops used AI “bots” to convert top secret intelligence down to a secret classification within seconds to make it easier to share with drone operators on the ground during the Iran war.

However, there is no doubt that AI also is helping the military find and strike targets.

The center that Toner oversees published a case study two years ago on how the Army’s 18th Airborne Corps used AI to target artillery strikes “just as efficiently as the best unit in recent American history” and with 2,000 fewer service members.

“Human operators are still the ones making crucial decisions, but AI … is making it possible to operate with a new level of speed and scale,” she said.

AI safety has created a public dispute between the Pentagon and Anthropic

The clash over the integration of AI into the military, who ultimately controls the technology and the ethics behind its use has played out in unusually public fashion during the Trump administration.

Hegseth and Anthropic are embroiled in a bitter contract dispute over the company’s concerns about unchecked government use of its technology, including the dangers of fully autonomous armed drones and of AI-assisted mass surveillance that could track dissent.

After CEO Dario Amodei refused to back down over concerns about how the chatbot Claude is used in classified Pentagon networks, both Trump and Hegseth accused Anthropic of endangering national security.

The Pentagon formally labeled the San Francisco-based company a supply chain risk — ending its $200 million defense contract and prohibited other government contractors from working with the company.

Anthropic sued, claiming the Pentagon is illegally retaliating by stigmatizing the company with a designation meant to protect against sabotage of national security systems by foreign adversaries. The Pentagon has since emphasized its turn to Anthropic rivals — including Google, OpenAI and SpaceX — to secure AI technology that can “augment warfighter decision-making in complex operational environments.”

Toner, a former OpenAI board member ousted after a clash with CEO Sam Altman, said “the general public often seems to underestimate the caution with which the U.S. military approaches new technologies.”

“Commanders want their missions to succeed, which means both being able to create lethal effects at scale, and avoiding unintended effects like friendly fire, civilian casualties, or simply identifying targets incorrectly,” she said.

LONDON (AP) — Samuel Smith spent years writing songs with a guitar in his hands.

Now, the London-based singer-songwriter is using artificial intelligence tools to help him continue making Americana music after Parkinson’s disease largely took away his ability to play guitar.

Smith, who was diagnosed with the progressive neurological disorder in 2020, recently released his second album, “The Art of Letting Go.” For one of the eight tracks, an instrumental piece titled “Horizon,” he relied on platforms that use AI to generate music to create demo arrangements that would convey his vision to the musicians who recorded the song.

The demos he created by humming rough melodies into his phone and uploading the recordings into song generators like Suno and Udio weren’t for mixing into the final studio version of “Horizon,” Smith stressed. But tremors, stiffness and fatigue, which are common symptoms of Parkinson’s, caused his guitar skills to deteriorate during the more than a year he worked on the album, he said.

“So then I’m faced with a question,” Smith, 49, said. “‘Don’t play, don’t be creative, or find a way out, find a route.’ And for me, this was the route.”

Generative AI has divided the music industry, whose artists and record labels have complained of their copyrighted work being used to train the models behind AI-powered music tools. Sony Music Entertainment, Universal Music Group and Warner Records sued Suno and Udio in June 2024, although Universal later reached a settlement and partnership deal with Udio and Warner did the same with Suno.

Less discussed is what those platforms can do when employed by a serious musician like Smith, whose disease affects the tools central to his songwriting and identity as a guitarist: his hands. He released his debut album, “In the Springtime,” in 2023, saying he wanted to give his two sons a way to remember when he could perform and record music himself.

“I’d always written, I’d also played, I always sung,” he said. “And immediately it became clear to me that I was in trouble, that my music was going to be seriously compromised.”

From prompts to convincing demos

AI music generators use systems trained on large datasets of recorded music and audio. The platforms analyze patterns in melody, harmony, and rhythm before generating new audio based on prompts or uploaded recordings. Users don’t need musical talent to end up with a serviceable song, or even a popular one.

Smith said producing convincing demos from the synthetic tracks the apps generated often required “50, 100, 150 attempts” and extensive editing “to get something that sounds close to my music.” After humming a song into his phone and uploading the recording, he gives prompts describing instrumentation, mood and style. .

“AI is not replacing anything for me,” he said. “It’s unlocking, it’s enabling. It’s allowing me to keep writing. I upload my lyrics; AI doesn’t create my lyrics. I upload my music; AI does not create my music.”

He added: “It then brings it to life in a way that I can play to session players and say, ‘Here, that’s what I’m thinking, that is what I’m hearing.’”

A bittersweet guitar duet

The album was produced by Grammy-winning pianist and producer Matt Rollings, who assembled a group of established roots and bluegrass musicians for the project. They included dobro player and 16-time Grammy winner Jerry Douglas, Grammy-winning banjo player Alison Brown, fiddler Stuart Duncan, guitarist Bryan Sutton, bassist Viktor Krauss and singers Jonatha Brooke and Glen Phillips.

For Smith, the experience of singing in a Nashville studio alongside musicians he had admired for decades was “an extraordinary moment.”

Grammy-nominated guitarist Julian Lage, known for his jazz and acoustic recordings with Blue Note Records, performed on the album’s title track and on “Horizon.” The latter recording became a bittersweet high point in Smith’s career; despite the progression of his disease, he managed to play a guitar duet with his friend.

“I hadn’t been able to play for months, but I kept telling myself that if I wrote something to take to the studio, perhaps the clouds would part for a few minutes,” Smith said. “That’s what happened. I had a window of about 10 minutes in the studio when my arm freed up. … So in the end, I was able to capture the last breath of my guitar playing.”

New possibilities and perils

Experts said AI-assisted music tools could benefit other people with disabilities or illnesses.

Ruaidhri Mannion, a composer, music producer and sonic artist who teaches at Brunel University of London, said technology like affordable digital recording software “effectively democratized the making of music” in recent decades. By helping songwriters and musicians communicate ideas and collaborate more easily, AI tools that generate polished-sounding material from voice or text prompts could work in the same way, he said.

“If these tools are able to enable people to be able to participate with other creative groups and encourage more people to feel confident to be able to reach out to an ensemble or an orchestra or something, then I think that is all for the better,” Mannion said.

But an overreliance on technology could intefere with the trial and error, frustration and synergy that are necessary parts of a musician’s artistic development, Mannion said.

“What makes a lot of music-making meaningful is the collaborative element,” he said. “There’s a lot of experimentation and development and failure that’s part of musical discovery.”

Udio and Suno have denied copyright infringement allegations and said they wanted to work with the music industry, not in opposition to it. Some musicians are unconvinced. A group of recording artists and activists, including singer-songwriter Tift Merritt, David Lowery of the bands Cracker and Camper Van Beethoven, and ECR Music Group President Blake Morgan, published an open letter in February under the heading “So no to Suno.”

“Many in our community are embracing responsible AI as a tool for creation, and as a means for fans to explore and interact with our artistry. That’s wonderful,” the letter read. “But it’s not the same as creating an environment where AI-generated works sourced from our music are mass distributed to dilute our royalties or, worse yet, reward those actively seeking to commit fraud. Artists need to know the difference.”

‘Show us what you can do’

Smith said he thinks his experience demonstrated how AI could benefit society and expand creative access, if it’s developed responsibly.

“My message would be that if these companies want to show they’ve got a place, a role in society, then step up,” Smith said. “Engage with health professionals, engage with music therapists, engage with society and show us what you can do.”

On May 21, Smith collaborated with the Berklee Music and Health Institute for an event in New York that brought together music industry leaders, researchers and clinicians to examine how music can support people living with neurological conditions. Smith discussed his experience living with Parkinson’s and sang again alongside musicians who played on “The Art of Letting Go.”

Creating music is crucial to the legacy Smith hopes to leave for his children, ages 4 and 17.

“My 4-year-old is probably never going to remember me playing, and it’s heartbreaking,” he said. “But I’ve been able to pull this into something and refuse to be defined by this disease.”

BUCHAREST, Romania (AP) — A Russian drone that was part of an overnight attack on Ukraine and then went astray slammed into an apartment building in eastern Romania, injuring two people in the NATO member country, Romanian authorities said, adding to concerns that the war could spread across the alliance’s borders.

The Russian drone was tracked by radar in Romanian airspace and crashed onto the roof of the building in the city of Galati, Romania’s Defense Ministry said in a statement. The impact was followed by a fire. The two people suffered minor injuries, and several others were evacuated.

The Romanian military scrambled two F-16 fighter jets and a helicopter that were authorized to engage targets, and alert messages were sent to residents of the affected areas. But the aircraft didn’t engage or shoot at the drone to bring it down.

Romania asked NATO for a faster transfer of anti-drone capabilities to its military, the Foreign Ministry said, calling the drone’s flight a serious violation of international law.

The incursion was the latest in a litany of drone incidents — from both Russia and Ukraine — to afflict NATO member states and leave the 32-member trans-Atlantic organization on edge, drawing strong condemnation from Romania’s allies.

Gen. Gheorghe Maxim, a stand-in commander for the Romanian armed forces’ joint staff, told a news conference on Friday that the strike in Galati wasn’t “an attack from Russia against Romania,” but “Romanians should understand that Russia is a threat to the security of the countries in the area.”

Galati is on the Danube River, near the borders of Ukraine and Moldova.

Ukrainian forces shot down 217 drones overnight on Friday, according to the country’s air force. In total, Russia attacked with 232 drones and one ballistic missile. Strikes were recorded in 14 areas, the air force said.

Romania calls it the worst of many incidents

While Romania has confirmed drone fragments on its territory on multiple occasions since the war started on Feb. 24, 2022, including in Galati in April this year, no one has previously been hurt in any of the drone incidents, in which many landed in remote areas.

In response to the latest incursion, Romanian President Nicusor Dan convened the NATO member’s top defense body for a meeting on Friday to discuss the implications of what he called “the worst incident to hit the national territory” since Russia launched a full-scale invasion of Ukraine, and blamed Russia directly for the incident.

“We will have proportional measures in relation to the Russian Federation. … There is no ambiguity about the author and the cause of this assault,” he wrote in a post on Facebook. He said that his thoughts are with the injured people, families and residents “who experienced terrible moments in their own homes.”

After the Supreme Council of National Defense meeting in Bucharest, Dan said that the Russian consul in the Black Sea port city of Constanta has been declared persona non grata and that the consulate there will be closed. Russia also has an embassy in Romania’s capital, Bucharest.

In recent years, airspace violations have become so common in Romania that lawmakers adopted legislation last year allowing the army to shoot down drones entering its airspace as a last resort. But Romania has remained cautious in downing errant drones, which can pose risks to populated areas.

Russia has been using long-range ballistic missiles and drones to damage Ukraine’s power grid and hammer cities, and Ukraine has braced for further heavy bombardments.

The latest incident adds to recent drone-related problems posed to Europe. Over the past few months, Ukrainian drones have crashed into the chimney of a power plant in Estonia, hit empty fuel tanks in Latvia and been shot down by Romanian fighter jets stationed in Lithuania. Ukrainian officials apologized and said that the drones were aimed at military targets inside Russia, but were sent off course by Russian electronic interference.

Since the war started nearly 4½ year ago, Poland, Croatia, Romania and non-NATO member Moldova have reported airspace violations and have found drone fragments on their territory.

The string of airspace violations has prompted questions about the state of air defenses on NATO’s eastern flank.

Allies rally to condemn incursion

NATO Secretary-General Mark Rutte said that he had spoken to Romania’s president and expressed the organization’s “absolute solidarity” with its ally.

Rutte said in an X post that he “affirmed that NATO stands ready to defend every inch of Allied territory. We will continue to enhance our readiness to deter and defend against any threat, including from drones.”

NATO allies are talking informally about the incursion, but no official meeting about it was due to take place on Friday. Romania can request formal NATO consultations, if it feels that it’s territory or security is under threat.

European Commission President Ursula von der Leyen also said the incident showed that Russia “has crossed yet another line.” She said the EU will keep strengthening security along its eastern border and was actively drafting another set of sanctions against Russia, the 21st so far.

“A Russian drone incursion struck a densely populated area in Romania, injuring civilians,” she wrote in a social media post. “On EU territory.”

Estonian Foreign Minister Margus Tsahkna said that the risk of such “serious incidents” was raised by “ Putin’s increasing nervousness, driven by military setbacks.”

___

Follow the AP’s coverage of the war in Ukraine at https://apnews.com/hub/russia-ukraine

___

This story has been corrected to remove the reference to Galati being east of the Ukraine and Moldova borders, since the city is to the west of them.

___

Stephen McGrath reported from Leamington Spa, England.

Artificial intelligence company Anthropic said Thursday it raised $65 billion in private funding that will push its valuation to $965 billion, a whopping number that makes the five-year-old maker of the Claude chatbot one of the world’s most valuable startups as it careens toward a likely Wall Street debut.

The announcement vaults Anthropic ahead of its chief rival, ChatGPT maker OpenAI, both in market value and in reported revenue. Anthropic said it’s now making annualized revenue of $47 billion from selling its technology to people and organizations using Claude to write code and do other work and personal tasks on their behalf.

Anthropic was formed in 2021 by ex-OpenAI leaders and now both AI firms, along with Elon Musk’s rocket and AI company SpaceX, are all expected to become publicly traded. All three are also still losing more money than they make, fueling concerns of an AI bubble.

San Francisco-based Anthropic said the new round of funding was led by investment firms Altimeter Capital, Dragoneer Investment Group, Greenoaks Capital and Sequoia Capital.

“This funding will help us serve the historic demand we are experiencing, stay at the research frontier, and bring Claude to more of the places where work happens,” said a written statement from Anthropic’s chief financial officer, Krishna Rao.

Anthropic also on Thursday launched its newest AI model, called Claude Opus 4.8, boasting that it is even better at coding and other professional work than previous models.

Anthropic’s meteoric rise and Claude’s growing popularity have left OpenAI playing catch-up despite its early lead in making ChatGPT a household name that sparked a commercial AI boom.

OpenAI last reported in March it was heading toward a $852 billion valuation after a $122 billion fundraising round. SpaceX was valued at $800 billion last year, but its value grew to $1.25 trillion after the space exploration company merged with Musk’s xAI in February. Musk recently announced plans for one of the biggest stock sales ever and will be able to pitch the offering to investors as soon as next week.

OpenAI also cleared a major hurdle toward its initial public offering ambitions after a federal court last week dismissed a lawsuit from Musk, an OpenAI co-founder and early donor, after a weeks-long jury trial over whether the company had betrayed its original nonprofit mission. Musk has said he plans to appeal.

Despite its newfound success, Anthropic has also faced obstacles this year — particularly a bruising legal fight with President Donald Trump’s administration over how AI tools like Claude can be used in warfare. Trump in February ordered all U.S. agencies to stop using Claude and Defense Secretary Pete Hegseth declared the company a supply chain risk after an unusually public clash between the Pentagon and CEO Dario Amodei. Anthropic sued in a dispute that is still working its way through two federal courts.

At the same time, Anthropic has been in talks with the White House over the cybersecurity capabilities and risks of its most powerful model, Mythos, which is not yet widely available to the public.

Anthropic also had an influential role at the Vatican ahead of Pope Leo XIV ‘s call Monday for robust regulation of AI and for its developers to work for the common good rather than profit.

The sweeping manifesto called “Magnifica Humanitas” (Magnificent Humanity), Leo’s first encyclical, repeatedly blasted the concentration of power and data in the hands of so few people in the private sector as a danger.

Ohio, one of the nation’s data center destination hot spots, is suspending a tax break that has been critical to its competition with other states to attract the massive new facilities that power and train artificial intelligence chatbots.

The move Wednesday by Republican Gov. Mike DeWine comes as tax breaks for energy-hungry AI data centers are increasingly playing a role in state budgets and the industry is under pressure to pay the full costs of the vast network of its computing warehouses needed to power AI.

The size of Ohio’s tax break skyrocketed, dwarfing previous projections, as opposition to data centers is sweeping through cities, suburbs and towns there and prompting lawmakers to form a committee to study the impact.

In the meantime, residents are trying to bypass the GOP-controlled Legislature and get a referendum on November’s midterm election ballot that’s designed to permanently ban hyperscale data centers, likely the strictest such statewide ban under consideration in the U.S.

DeWine’s office cited the rising utilization of the tax break and the state Legislature’s new research undertaking to declare a “pause” in granting it to new applicants.

“The governor felt it was the right time to let the citizens know, let businesses know that we’re going to pause on new offers of this tax incentive while that process plays out,” DeWine’s spokesperson, Dan Tierney, said Thursday.

DeWine stressed that he supports data centers — calling them a critical component in today’s economy — and that the roughly $37 billion in data center-related investment in 2024 and 2025 in Ohio has been worthwhile. Meanwhile, business groups — including the state Chamber of Commerce — and labor unions warned that pausing the tax break put Ohio at risk of losing tech-sector investments to other states.

The state, in 2024, had used previous history in projecting that the exemption would total $136 million in fiscal 2025 and $142 million in fiscal 2026. It was $554 million in 2024 and nearly $1.6 billion in 2025, the state reported.

The resumption of Ohio’s tax break — should it resume — could happen under a new governor: DeWine is term-limited and the race is on to replace him. The Republican nominee, Republican Vivek Ramaswamy — an Ivy League-educated biotech billionaire — likes to talk about turning the Ohio River Valley into the next Silicon Valley.

However, Ramaswamy and Democratic nominee Amy Acton could share the midterm ballot in November with the citizen-led drive to ban the new construction of large data centers across Ohio. It faces a July 1 deadline to gather more than 400,000 voter signatures.

State tax breaks for the massive data center industry are facing growing criticism by governors and lawmakers.

The cost is likely rising as data center and AI-related investments drive higher consumer spending in the U.S. and tech giants keep boosting their spending commitment to hyperscale data centers.

In Virginia, budget negotiations have been hung up for months on a bid by Senate Democrats to eliminate the roughly $1.6 billion annual tax break.

Thirty-eight states have some form of a sales tax break for data centers, according to the National Conference of State Legislatures.

Many were approved when data centers were a small, but growing part of the economy, and well before the late 2022 debut of OpenAI’s ChatGPT launched an intensifying buildout of increasingly large data centers.

Ohio’s exemption is fairly broad, applying not only to construction materials, but to the expensive equipment — such as server racks and cooling systems — used in data centers. Operators might buy new server racks every couple of years as the technology improves.

DeWine’s order was a surprise.

Dorsey Hager, executive secretary-treasurer of the Columbus/Central Ohio Building and Construction Trades Council, where union members spend much of their time on data center projects, said he was upset with DeWine and trying to understand the governor’s reasons.

He worried, he said, that developers that were in the midst of trying to finalize plans or permits for a project might have second thoughts.

Lawmakers acknowledged the opposition in announcing their joint data center committee on May 13, and said their mission was to ensure that Ohioans have accurate information about the economic, environmental and security impact of data center development.

“We’re well aware of initiatives to limit Ohio data center development during this critical point in America’s history,” state Rep. Adam Holmes told a news conference. “This public concern has become a priority issue for us and could have dramatic impact on Ohio and American’s future.”

___

Follow Marc Levy at http://twitter.com/timelywriter

NEW YORK (AP) — U.S. prosecutors slapped insider trading charges against a Google employee this week, alleging the software engineer used confidential company information to pocket more than $1.2 million from prediction market platform Polymarket with bets on search trends.

In a complaint unsealed in New York, authorities identified the employee as 36-year-old Michele Spagnuolo — an Italian citizen residing in Switzerland who has worked for Google since 2014. Under the online name “AlphaRaccoon,” they alleged, Spagnuolo used the company’s 2025 “Year in Search” data before it was published to enter Polymarket wagers about the most trending Googled people of last year.

This week’s charges “reinforce a decades-old message: corporate insiders cannot use confidential business information to turn a profit in our markets,” Jay Clayton, U.S. Attorney for the Southern District of New York, said Wednesday. “Insider trading compromises the integrity of our markets, and the American people want this greed-driven conduct investigated and prosecuted.”

Spagnuolo allegedly made new Polymarket trades as Google’s internal search data evolved, from October into December of last year. For example, per the complaint, Spagnuolo initially wagered that Kendrick Lamar — who headlined the 2025 Super Bowl halftime show — would top search trends for people last year. But after internal Google data showed that alt-pop singer D4vd was later leading the influx of searches, he placed new bets. D4vd, whose legal name is David Burke, was charged last month with murdering 14-year-old Celeste Rivas Hernandez.

Using the prediction market’s “yes” or “no” wagers, Spagnuolo also made a series of Polymarket trades about other individuals who would or wouldn’t rank in Google’s 2025 search trends, the complaint said. And after the data was published on Dec. 4, the AlphaRaccoon account soon pocketed sizeable profits. An FBI investigation later traced its cryptocurrency payments.

An attorney for Spagnuolo was not immediately identified. California-based Google confirmed to The Associated Press it had placed its employee on leave.

“The employee accessed our marketing material using a tool available to all employees, but using such confidential information to place bets is a serious breach of our policies,” a Google spokesperson said in a statement — adding the company was working with law enforcement and “will take the appropriate action.”

Polymarket reiterated it too worked closely with authorities. A spokesperson also touted that the company “is the only prediction platform to date whose cooperation has led to insider trading charges in the United States” — and maintained blockchain trading, which Polymarket uses, is “transparent, traceable, and bad actors leave footprints.”

Spagnuolo isn’t first person to face insider trading charges spanning from Polymarket trades. Last month, the government also charged a special forces soldier who made over $400,000 from Polymarket trades betting on the downfall of former Venezuelan President Nicolás Maduro. The solider allegedly used classified information ahead of January’s U.S. military operation, which he was a part of.

Such scandals have put the spotlight on a murky (and growing) world of speculative, 24/7 transactions now filling the internet. Prediction markets sell event contracts — so they’re also categorized and regulated differently from traditional forms of gambling. That’s raised concerns about consumer protections, and legal battles over government oversight.

President Donald Trump’s administration has already thrown its support behind company operators — and sued several states over their regulation efforts. Meanwhile, the industry is scrambling to assure the public with new guardrails. Polymarket recently rewrote its rules to clearly state users cannot trade on contracts where they might possess confidential information, or could influence the outcome of an event.

Spagnuolo is being charged with violating the U.S. Commodity Exchange Act, wire fraud and money laundering. He could face years of prison time.

LONDON (AP) — Temu was hit with a 200 million euro ($232 million) fine Thursday after a European Union investigation found the Chinese online retailer failed to protect consumers from illegal products like toxic or hazardous toys and unsafe electronics.

The 27-nation EU’s fine follows preliminary findings last year that Temu was exposing consumers to a high risk of products sold on its platform like baby toys and small electronics that didn’t comply with EU consumer safety rules.

The bloc’s executive arm issued the penalty under the Digital Services Act, or DSA, a wide-ranging rulebook that requires online platforms to do more to keep internet users safe from harmful content or dodgy goods, under the threat of hefty fines.

It’s the second time Brussels has issued a fine under three-year-old DSA, following a $120 million penalty last year for Elon Musk’s social media site X.

Temu said it disagreed with the decision and considered the fine “disproportionate.”

The decision relates to the commission’s first DSA evaluation of Temu in 2024 “and does not reflect the current state of our systems,” the company said.

“Temu engaged constructively with the Commission throughout the process and has since taken further steps to strengthen risk assessment, platform governance, and user protection,” it said in a statement.

The company is popular because it offers cheap goods – from clothing to home products — shipped from sellers in China. The platform has 92 million users in the EU and is owned by PDD Holdings Inc., which also owns the popular Chinese e-commerce site Pinduoduo.

The European Commission said Temu failed to identify, analyze and assess the systemic risks of illegal goods for sale on the platform and the resulting harm to European consumers.

Investigators had carried out a “mystery shopping exercise” that turned up a number of “non-compliant” products, including many electronic device chargers that failed basic safety tests. They also found a very high percentage of baby toys that posed safety risks, either because they contained chemicals at levels that exceeded safety limits or because they had parts that came off and could be a suffocation risk.

The commission said failing to do proper risk assessments is a particularly serious breach of the bloc’s digital rules.

Risk assessments are “not box‐ticking exercises,” European Commission Executive Vice-President Henna Virkunnen said.

“Temu’s risk assessment underestimates concrete risks, lacks specificity, is not grounded in solid evidence, and is not comprehensive,” she said in a prepared statement. “It leaves regulators, users, and the public in the dark about the true scale of potential harm posed by illegal products sold on Temu. Now it is time for Temu to comply with the law.”

Temu has until the end of August to submit an “action plan” to remedy the problem. It could be hit with additional daily, weekly or monthly fines if it fails to comply.

___

AP writer Sam McNeil in Brussels contributed to this report.

TOKYO (AP) — Mechanical hands dexterous enough to thread a needle, childlike dancing robots and adult-sized ones to help with deliveries were on display Thursday as the Humanoids Summit Tokyo opened.

Among the dozens of companies taking part, including well-known players like Boston Dynamics and Toyota Motor Corp., the big stars now were clearly the Chinese.

Chinese newcomers, like Booster Robotics and LimX Dynamics, took the technology initially developed in Japan and the U.S. and fine-tuned it, often for cheaper mass production. It’s a repeat of what happened in other Japanese industries, from consumer electronics to cellphones and electric vehicles. In humanoids, Japan was initially ahead but then failed to produce major commercial solutions.

Tim Hornyak, author of “Loving the Machine: The Art and Science of Japanese Robots,” who was at the event, categorized it as the so-called “Galapagos syndrome,” referring to how innovative Japanese products evolve in isolation and end up not translating for the international market.

“I really hope that Japan can come up with a Ford Model T-version of humanoid roots. But I think China has already stolen their lunch. It’s a bit too little too late,” he said.

The dancing and wiggling Mini Pi Plus robot from High Torque of China, for instance, still can’t help at an auto plant or do your dishes. But it’s cute. And it doesn’t come with an eye-popping price tag, starting at $5,500.

Chinese robots are dominating

One telling example of Chinese robotics use in Japan was GMO, a Tokyo-based AI and robotics company working on a humanoid with camera eyes that will help with Japan Airlines cargo and other chores at an airport.

The key is to have the robot do the work in the same way as people so they would be interchangeable, an initiative meant to tackle the labor shortage problem that is increasingly serious in Japan.

The inner robotics workings were all courtesy of Unitree, a Chinese outfit, which is also working on a four-legged dog-like “stellar explorer.”

Experts say Japan, with its finesse in manufacturing, proved a good breeding ground for robotics development. The sociological backdrop of a public receptive to robotics also helped.

A recent Pew global survey showed that people in Japan are highly aware of AI but are less anxious about it, at about 28%, than people in the U.S. at 50%.

Japanese automaker Honda Motor Co., a leader in robotics with its walking humanoid Asimo, first shown in 2000, was demonstrating a motorized four-fingered robotic hand that could screw on and off tiny bolts, or thread a needle.

It didn’t seem to bother Keisuke Tsuta, assistant chief engineer, that similar mechanical hands were on display galore near his booth, many of them from Chinese makers.

Japanese robotics show their prowess

The technology Honda had developed is more durable and powerful than rival offerings, and the Japanese have historically shown they can excel at quality mass production, according to Tsuta.

The looming threat of a Chinese robotics domination didn’t seem to phase Osaka University Professor Hiroshi Ishiguro, who has worked on humanoids for decades, including one that’s his clone.

“What’s significant is that Japan has a culture that’s receptive to robotics. If we’re going to really start using robots in society, Japan is the ideal place,” he said, stressing that Japanese don’t discriminate against robots.

His robotic counterpart, dressed all in black like the professor, did as good a job, if not better, of answering a key existentialist question on the meaning of robots.

“I think robots will coexist with people. Robots are the mirror of human beings,” the robot replied in a slightly monotonous but human-like voice.

Earlier, the professor had answered a similar question, but a bit differently.

“No one is interested in me. All everyone cares about is my robot,” he said, sitting next to his twin-like humanoid.

“As long as people identify with what I have produced, I am a success,” he added. ___

Yuri Kageyama is on Threads: https://www.threads.com/@yurikageyama

TORONTO (AP) — Canada will buy early warning radar planes built by Sweden’s Saab and Canada’s Bombardier over two American options, Prime Minister Mark Carney announced Wednesday.

Carney said his government has entered negotiations to procure Saab’s Airborne Early Warning & Control Aircraft, which is built on the Canadian-manufactured Bombardier Global 6500 aircraft and will support domestic production.

Carney also noted it is made with 20% U.S. content. The federal government has previously said it’s in the market for six radar aircraft.

Carney has made a point of diversifying its military spending away from the United States.

Equipped with powerful radar, the Saab planes provide situational awareness about aircraft and missile movements for hundreds of miles (kilometers). They can detect hostile activity in the air or from ships are able to direct fighter jets to their targets.

“Saab’s GlobalEye will be a key resource for the Canadian Armed Forces to detect and deter threats across the Arctic,” Carney said.

The other options would have been to buy the E-7A Wedgetail surveillance planes built by U.S. aircraft manufacturer Boeing or the Aeris X by L3Harris.

NATO is also considering the Saab plane over the American options.

Carney has previously said that no more will over 70 cents of every dollar of Canadian military capital spending go to the U.S.

U.S. President Donald Trump’s actions — including launching a trade war and suggesting Canada become the 51st U.S. state — infuriated Canadians and created the political environment for Carney to win the job of prime minister after promising to confront Trump’s increased aggression.

The Canadian government is reviewing the planned purchase of U.S. F-35 fighter jets to explore other options. Carney has said the potential for having more production in Canada is a factor. A proposal by Saab promised that assembly and maintenance of the Saab Gripen fighter jet would take place in Canada.

Swedish Prime Minister Ulf Kristersson said Canada’s decision on the early warning radar plane ties the two nations even closer together.

“As the GlobalEye is based on the Canadian Bombardier Global 6500 platform, GlobalEye is already creating jobs in Canada, and working with the Canadian supply chain,” Kristersson said in a post on social media.

DURHAM, N.C. (AP) — A robot being developed at Duke University is almost ready to face the world, in any direction.

Instead of trying to copy symmetrical shapes from nature by building robots that look like people, dogs or insects, engineering professor Boyuan Chen and his team focused on uniformity in action, or what he calls “dynamic symmetry.”

The result was Argus. The roly-poly robot named after a mythological many-eyed giant has depth-sensing cameras attached to 20 telescoping legs that radiate from a central core. With no front, back, top or bottom, it can see and move in any direction instantly.

“Instead of measuring how your legs are arranged around a different part of your body, we’re measuring how fast you can move in any direction,” Chen said. “Who said, you know, if you have a robot to help us in a most effective way, it has to look like us?”

In experiments, Argus has navigated sandy beaches and forest undergrowth, rolling over obstacles and stabilizing itself after being pushed. It can climb between parallel brick walls by alternating bracing and thrusting motions with its legs. If one or more motor dies or a leg breaks, it continues to function.

“Watching Argus move is unlike watching any other robot we’ve worked with,” said Jiaxun Liu, a graduate student and co-author of a study about Argus published online Wednesday in the journal Science Robotics. “The first time we saw it navigate among trees and rough terrain, even under heavy collisions, we knew this was something different.”

As part of their work, researchers developed a new design principle called dynamic isotropy that rates robots on a scale of 0 to 1 based on how uniformly they can accelerate in every direction. Most robots in use today, including humanoids and drones, score below 0.6. Argus scores 0.91.

“When a robot can accelerate equally well in every direction, it stops needing to face the world in any particular way,” said Chen, who hopes the same principle could guide the development of search and rescue robots, underwater or aerial vehicles or robots with the ability to grip objects.

“Instead of building a robot hand that looks like a human hand … one idea is to think about having Argus be the hand itself, and it can manipulate objects in any direction,” he said. “The knowledge we can transfer to the rest of the world is much more deeper than building an existing robot or copying an existing species.”

____

Ramer reported from Concord, New Hampshire.