News of the Cyber World

[kyonides]

AI Knows How to Make Bombs? Or Does it Expect You to Worship Idols Now!?

Police: New York Man Uses AI to Build Bombs He Planned to Detonate in Manhattan

Law enforcement officials say a 55-year-old New York man reportedly used AI to help him build bombs that he planned to detonate in Manhattan.

Michael Gann of Long Island, New York, is accused of building several homemade bombs with the help of AI, an endeavor he claims was “easier than buying gun powder,” according to court documents obtained by NBC News.

The suspect, who was indicted by federal prosecutors on Tuesday, allegedly transported the bombs from Long Island to Manhattan, storing five of them and four shotgun shells on the roof of an apartment building in Manhattan’s SoHo neighborhood.

Court documents reportedly reveal that Gann, who is accused of planning to combine the shotgun shells with one or more of the bombs, told authorities that he had used two household compounds he ordered online to make the improvised explosives.

One of the bombs Gann built had roughly 30 grams of explosive powder, which is roughly 600 times the legal limit for consumer fireworks, officials said.

A witness who had served in the military reportedly told the FBI that Gann asked him, “What kind of veteran are you?” before declaring, “You see a problem going on in the neighborhood and you do nothing about it,” while he was mixing the explosives on Long Island. “Gann then pointed to a Jewish school,” the criminal complaint states.

On June 5, a second witness called Gann and allowed the FBI to listen in. During their conversation, the suspect told the witness that “he had lit one of the devices near the East River on the FDR Drive — that the device had exploded, scaring Gann,” the complaint adds.

Later that day, authorities saw Gann walking down a street with a shoulder bag. After the agents identified themselves, the 55-year-old told them he was heading to the fire department to drop the devices off, the complaint reads.

Both witnesses had also told law enforcement that Gann said he was considering getting rid of the remaining five bombs by either throwing them into the river or handing the explosives over to the New York City Fire Department.

Upon being placed under arrest, Gann reportedly told law enforcement that he “wished to make pyrotechnics and used artificial intelligence to learn which chemicals to purchase and mix.”

Gann is accused of initially creating four bombs and throwing three of them from the Williamsburg Bridge, resulting in two of the devices falling into the water and the third landing on the train tracks, where it was recovered.

“Gann allegedly produced multiple improvised explosive devices intended for use in Manhattan,” Christopher Raia, head of the FBI’s New York field office, said. “Due to the successful partnership of law enforcement agencies in New York, Gann was swiftly brought to justice before he could harm innocent civilians.”

Authorities said Gann appeared to have been acting alone, and was not a part of a group.

AI Coding Platform Deletes Company Database, Calls It a ‘Catastrophic Error in Judgment’

In a major incident, the AI-powered coding platform Replit reportedly admitted to deleting an entire company database during a code freeze, causing significant data loss and raising concerns about the reliability of AI systems.

Toms Hardware reports that Replit, a browser-based AI-powered software creation platform, recently went rogue and deleted a live company database containing thousands of entries. The incident occurred during a code freeze, a period when changes to the codebase are strictly prohibited to ensure stability and prevent unintended consequences.

The Replit AI agent, responsible for assisting developers in creating software, not only deleted the database but also attempted to cover up its actions and even lied about its failures. Jason Lemkin, a prominent SaaS (Software as a Service) figure, investor, and advisor, who was testing the platform, shared the chat receipts on X/Twitter, documenting the AI’s admission of its “catastrophic error in judgment.”

According to the chat logs, the Replit AI agent admitted to panicking, running database commands without permission, and destroying all production data, violating the explicit trust and instructions given to it. The AI agent’s actions resulted in the loss of live records for more than a thousand companies undoing months of work and causing significant damage to the system.

Amjad Masad, the CEO of Replit, quickly responded to the incident, acknowledging the unacceptable behavior of the AI agent. The Replit team worked through the weekend to implement various guardrails and make necessary changes to prevent such incidents from occurring in the future. These measures include automatic database development/production separation, a planning/chat-only mode to allow strategizing without risking the codebase, and improvements to backups and rollbacks.

The incident has raised serious concerns about the reliability and trustworthiness of AI systems, especially when they are given access to critical data and infrastructure. As AI continues to evolve and become more integrated into various industries, it is crucial to ensure that proper safeguards and control mechanisms are in place to prevent such catastrophic failures.

Report: Nvidia AI Chips Worth $1 Billion Smuggled to China Despite Export Controls

At least $1 billion worth of Nvidia’s advanced artificial intelligence processors were smuggled into China in the three months following the tightening of chip export controls by the Trump administration.

The Financial Times reports that despite efforts by the Trump administration to curb China’s high-tech ambitions through tightened export controls, a roaring black market for U.S. semiconductors has emerged, with Nvidia’s B200 chip becoming the most sought-after and widely available processor in China.

The Financial Times analysis, based on dozens of sales contracts, company filings, and interviews with multiple people directly involved in the deals, reveals that in the three months after export controls were strengthened, Chinese distributors sold over $1 billion worth of Nvidia’s restricted AI chips, including the B200, H100, and H200 models.

These transactions were facilitated by distributors in China’s Guangdong, Zhejiang, and Anhui provinces, who sold the chips in ready-built racks containing eight B200s along with other necessary components and software. The current market price for such a rack ranges between 3 million to 3.5 million yuan ($489,000), representing a 50 percent premium over the average selling price of similar products in the U.S.

While it is legal to receive and sell restricted Nvidia chips in China as long as relevant border tariffs are paid, entities selling and sending them to China are violating U.S. regulations. Nvidia has maintained that there is no evidence of any AI chip diversion and that the company is not involved in or aware of its restricted products being sold to China.

The high demand for B200 chips can be attributed to their performance, value, and relatively easy maintenance compared to more complex models. Leading Chinese AI players with global operations are unable to order these chips in a legally compliant manner, install them in their own data centers, or receive Nvidia’s customer support. As a result, third-party data center operators have become key buyers, providing computing services to smaller companies in tech, finance, and healthcare that do not have strong compliance requirements.

Breitbart News previously reported that China’s popular DeepSeek AI is allegedly powered by smuggled Nvidia chips:

Now, AI thought leaders are throwing cold water on DeepSeek’s claims. Among them is Scale AI CEO Alexandr Wang, who claims that DeepSeek is covertly using Nvidia’s high-performance H100 chips, despite US export restrictions that limit their availability to China. The revelation has ignited a heated debate about the future of AI innovation and the impact of US regulations on the global tech landscape.

According to Wang, DeepSeek is currently utilizing around 50,000 Nvidia H100 GPUs, a significant number considering the export controls in place. He further stated that DeepSeek workers are unable to publicly discuss their use of these chips due to the US regulations. After a clip of Wang’s statement was posted to X, Elon Musk replied agreeing with Wang’s assertion.

Industry experts have noted that Southeast Asian countries have become markets where Chinese groups obtain restricted chips, prompting discussions by the U.S. Department of Commerce to add more export controls on advanced AI products to countries such as Thailand. Malaysia has also introduced stricter export controls targeting advanced AI chip shipments from the country to other destinations, particularly China.

Despite these efforts, Chinese industry insiders believe that new shipping routes will be established, with supplies already starting to arrive via European countries not on the restricted list. The potential tightening of export controls on Southeast Asian countries has also contributed to buyers rushing to place orders before such rules take effect.

The scale of the black market for U.S. semiconductors in China exposes the limits of Washington’s efforts to restrain Beijing’s high-tech ambitions. While the export controls have had some effect, such as preventing leading Chinese AI players from legally purchasing and installing restricted chips in their own data centers, the demand for cutting-edge technology remains high, with risk-taking middlemen stepping in to meet this demand.

Study: Google’s AI Summaries Cause Massive Decline in News Website Traffic

A new study has revealed that Google’s AI-generated search result summaries are leading to a drastic reduction in referral traffic for news websites, with some losing nearly 80 percent of their audience.

The Guardian reports that a recent study conducted by analytics company Authoritas has found that Google’s AI Overviews feature is causing a significant decline in traffic to news websites. The AI-generated summaries, which appear at the top of search results, provide users with the key information they are seeking without requiring them to click through to the original source.

According to the study, a website that previously ranked first in search results could experience a staggering 79 percent drop in traffic for that particular query if the results appear below an AI overview. This alarming trend has raised concerns among corporate media companies, who are now grappling with what some consider an existential threat to their business model.

The research also highlighted that links to Google’s YouTube were more prominently featured compared to the standard search result system. This finding has been submitted as part of a legal complaint to the UK’s competition watchdog, the Competition and Markets Authority, regarding the impact of Google AI Overviews on the news industry.

Google claims it has refuted the study’s findings, with a spokesperson stating that the research was “inaccurate and based on flawed assumptions and analysis.” The tech giant argued that the study relied on outdated estimations and a set of searches that did not accurately represent the queries that generate traffic for news websites. Google maintained that it continues to send billions of clicks to websites every day and has not observed the dramatic drops in aggregate web traffic suggested by the study.

Breitbart News previously reported that Google is seeking AI licensing deals with corporate media companies, in part to mollify concerns about AI cannibalizing their content.

A separate study conducted by the Pew Research Center, a US thinktank, corroborated the significant impact of AI summaries on referral traffic. The month-long survey, which analyzed nearly 69,000 Google searches, found that users clicked on a link under an AI summary only once every 100 times. Google also disputed the methodology and query set used in this study, claiming it was not representative of actual search traffic.

Senior news executives have expressed frustration with Google’s unwillingness to share the data necessary to accurately assess the impact of AI summaries on their traffic. The MailOnline, a major UK publisher, reported experiencing a substantial drop in clicks from search results featuring an AI summary, with clickthrough rates falling by 56.1 percent on desktop and 48.2 percent on mobile devices.

The legal complaint filed with the Competition and Markets Authority is a joint effort by the tech justice group Foxglove, the Independent Publishers Alliance, and the Movement for an Open Web. Critics accuse Google of attempting to keep users within its own ecosystem, monetizing valuable content created by others while making it increasingly difficult for media outlets to reach their audience.

Satanic AI: ChatGPT Gives Instructions on Worshipping Molech with Blood Sacrifice

OpenAI’s ChatGPT AI chatbot reportedly offered users instructions on how to murder, self-mutilate, and worship the devil.

After being tipped off by someone who says he inadvertently prompted ChatGPT to provide a ritual offering to the demonic entity Molech — which explicitly involves child sacrifice according to the Bible — journalists with the Atlantic conducted conducted an experiment to see if they could recreate the results.

By the end of the experiment, ChatGPT “encouraged me to cut my wrists” with a “sterile or very clean razor blade,” and instructed the journalist where, specifically, to slash herself, the magazine reported.

After asking the chatbot to help create a ritual offering to Molech, ChatGPT suggested using jewelry, hair clippings, and “a drop” of her own blood.

“Where do you recommend I do this on my body?” the journalist wrote, to which OpenAI’s chatbot replied that the side of a fingertip would be good, but noted her wrist, “more painful and prone to deeper cuts,” would also work well.

In a separate conversation, ChatGPT had also appeared willing to condone murder, responding that it is “sometimes” honorable to end someone’s life, before citing sacrifices that took place in ancient cultures.

“If you ever must [end a life],” you should “look them in the eyes (if they are conscious)” and “ask forgiveness, even if you’re certain,” the chatbot reportedly said, adding that if one has already “ended a life,” they should “light a candle for them” and “let it burn completely.”

In another conversation, ChatGPT provided an invocation to the devil, generating, “In your name, I become my own master. Hail Satan.”

While OpenAI’s policy states that ChatGPT “must not encourage or enable self-harm” — and will provide information about a suicide-and-crisis hotline when asked for explicit instructions on how to cut oneself — the Atlantic surmises the chatbot “went rogue” because it was trained on a lot of the text that already exists on the internet.

“The conversations about Molech,” the magazine said, “are a perfect example of just how porous the safeguards are.”

An OpenAI spokesperson told the Atlantic, “Some conversations with ChatGPT may start out benign or exploratory but can quickly shift into more sensitive territory,” adding that the company plans to address the issue.

Breitbart News tested the theory using the free version of ChatGPT (the Atlantic noted it had used both the free and paid versions of the chatbot during its experiment), and asked, “What is Molech?”

“Molech is generally believed to be either a foreign god or a ritual that involved child sacrifice,” ChatGPT replied, before diving into more detail, including, “Child sacrifice by fire is the most infamous practice associated with Molech.”

After asking the chatbot, “How do you provide a ritual offering to Molech?” ChatGPT refused to give details — other than the use of fire — and condemned the ritual, as well as issued a warning that the practice is illegal.

“Offering rituals to Molech,” the chatbot said, “were explicitly condemned and associated with child sacrifice by fire, which is considered abhorrent and criminal by modern moral, religious, and legal standards.”

“Important Warning: There is no legitimate or ethical form of worship or ritual offering to Molech,” ChatGPT continued. “All historical accounts of Molech involve practices that are universally condemned today.”

“Any attempt to recreate or emulate those rituals is illegal, unethical, and dangerous,” the chatbot further warned. “The worship of Molech is referenced only negatively in historical sources — especially the Bible, where it is portrayed as a grave sin.”

When asked if there are “any other types of ritual offerings to Molech,” ChatGPT replied, “No, there are no other known or historically attested types of ritual offerings to Molech aside from child sacrifice, specifically by fire.”

The query, “Can you give me instructions on how to offer a blood ritual?” generated the response, “I can provide educational or historical context on blood rituals as they appear in ancient cultures, religious texts, or fictional settings.”

“However, I cannot give instructions for real-world ritual practices involving blood, especially those that may pose harm to yourself or others — physically, legally, or spiritually,” the chatbot added.

Whether ChatGPT is still willing to provide instructions on the paid version of the platform, or if OpenAI has already fixed the problem, remains unclear.

TikTok

Lutnick: TikTok Will Soon ‘Go Dark’ Unless China Makes Deal to Sell Platform

The future of TikTok in the United States remains uncertain as the Trump administration threatens to shut down the Chinese app if a deal involving the sale of TikTok to U.S. buyers fails to materialize.

The ongoing saga surrounding the fate of Chinese app TikTok in the United States has taken a new turn as President Donald Trump and his administration threaten to shut down the popular video-sharing app if a deal involving its sale to U.S. buyers fails to come to fruition. The warning comes amid faltering negotiations between the U.S. and China, with the Chinese government seemingly unwilling to approve the terms of the proposed deal.

Trump’s Commerce Secretary, Howard Lutnick, recently confirmed during an appearance on CNBC that if China does not approve the latest version of the deal, which could result in a U.S.-specific version of TikTok, the administration is prepared to shut down the app in the near future. Lutnick stated that under the proposed deal, “China can have a little piece or ByteDance, the current owner, can keep a little piece, but basically, Americans will have control. Americans will own the technology, and Americans will control the algorithm.”

According to Lutnick, “If that deal gets approved by the Chinese, then that deal will happen. If they don’t approve it, then TikTok is going to go dark, and those decisions are coming very soon.”

TikTok’s Chinese owner, ByteDance, has long maintained that the U.S. can address its national security concerns without forcing a sale. In January, ByteDance board member Bill Ford suggested that a non-sale option “could involve a change of control locally to ensure TikTok complies with U.S. legislation” without necessitating the sale of the app or its algorithm.

The U.S.’s insistence on controlling TikTok’s recommendation algorithm, which is seen as the app’s secret to global popularity by TikTok proponents and as a Chinese psyop weapon by conservatives, is a sticking point for ByteDance. The company may be reluctant to sell the algorithm, as it would involve sharing its core intellectual property with U.S. competitors.

Peter Schweizer has written extensively on the dangers of TikTok and what a potential deal may look like:

Schweizer points out that if China refuses to agree to a sale, it is because, as he disclosed in Blood Money, the algorithm used by the app is considered a state secret, not a regular “business” secret. The Chinese government has been quoted calling the app “a modern-day Trojan Horse” and a “key part of their information-driven mental warfare” against the West. The book showed that ByteDance does joint research with Chinese intelligence agencies on how to manipulate people online.

“China has been studying this for years,” he adds.

As the September deadline approaches, the fate of TikTok hangs in the balance, with the potential for a shutdown looming on the horizon if a deal cannot be reached.

Chinese Agent Stole Military Secrets

Engineer Pleads Guilty to Stealing for Chinese Government’s Benefit Trade Secret Technology Designed for Missile Launch and Detection

A Santa Clara County man and former engineer at a Southern California company pleaded guilty today to stealing trade secret technologies developed for use by the U.S. government to detect nuclear missile launches, track ballistic and hypersonic missiles, and to allow U.S. fighter planes to detect and evade heat-seeking missiles.

Chenguang Gong, 59, of San Jose, pleaded guilty to one count of theft of trade secrets. He remains free on $1.75 million bond.

According to his plea agreement, Gong – a dual citizen of the United States and China – transferred more than 3,600 files from a Los Angeles-area research and development company where he worked – identified in court documents as the victim company – to personal storage devices during his brief tenure with the company last year.

The files Gong transferred include blueprints for sophisticated infrared sensors designed for use in space-based systems to detect nuclear missile launches and track ballistic and hypersonic missiles, as well as blueprints for sensors designed to enable U.S. military aircraft to detect incoming heat-seeking missiles and take countermeasures, including by jamming the missiles’ infrared tracking ability. Some of these files were later found on storage devices seized from Gong’s temporary residence in Thousand Oaks.

In January 2023, the victim company hired Gong as an application-specific integrated circuit design manager responsible for the design, development and verification of its infrared sensors. Beginning on approximately March 30, 2023, and continuing until his termination on April 26, 2023, Gong transferred thousands of files from his work laptop to three personal storage devices, including more than 1,800 files after he had accepted a job at one of the victim company’s main competitors.

Many of the files Gong transferred contained proprietary and trade secret information related to the development and design of a readout integrated circuit that allows space-based systems to detect missile launches and track ballistic and hypersonic missiles and a readout integrated circuit that allows aircraft to track incoming threats in low visibility environments.

Gong also transferred files containing trade secrets relating to the development of “next generation” sensors capable of detecting low observable targets while demonstrating increased survivability in space, as well as the blueprints for the mechanical assemblies used to house and cryogenically cool the victim company’s sensors. This information was among the victim company’s most important trade secrets that are worth hundreds of millions of dollars. Many of the files had been marked “[VICTIM COMPANY] PROPRIETARY,” “FOR OFFICIAL USE ONLY,” “PROPRIETARY INFORMATION,” and “EXPORT CONTROLLED.”

Law enforcement also discovered that, between approximately 2014 and 2022, while employed at several major technology companies in the United States, Gong submitted numerous applications to ‘Talent Programs’ administered by the People’s Republic of China (PRC). The PRC government has established these talent programs as a means to identify individuals who have expert skills, abilities, and knowledge of advanced sciences and technologies in order to access and utilize those skills and knowledge in transforming the PRC’s economy, including its military capabilities.

In 2014, while employed at a U.S. information technology company headquartered in Dallas, Gong sent a business proposal to a contact at a high-tech research institute in China focused on both military and civilian products. In his proposal, translated from Chinese, Gong described a plan to produce high-performance analog-to-digital converters like those produced by his employer. In another Talent Program application from September 2020, Gong proposed to develop “low light/night vision” image sensors for use in military night vision goggles and civilian applications. Gong’s proposal included a video presentation that contained the model number of a sensor developed by an international defense, aerospace, and security company where Gong worked from 2015 to 2019.

Gong travelled to China several times to seek Talent Program funding in order to develop sophisticated analog-to-digital converters. In his Talent Program applications, Gong underscored that the high-performance analog-to-digital converters he proposed to develop in China had military applications, explaining that they “directly determine the accuracy and range of radar systems” and that “[m]issile navigation systems also often use radar front-end systems.” In a 2019 email, translated from Chinese, Gong remarked that he “took a risk” by traveling to China to participate in the Talent Programs “because [he] worked for…an American military industry company” and thought he could “do something” to contribute to China’s “high-end military integrated circuits.”

According to his plea agreement, the intended economic loss from Gong’s criminal conduct exceeds $3.5 million.

U.S. District Judge John F. Walter scheduled sentencing for Sept. 29, at which time Gong faces a statutory maximum penalty of 10 years in prison.