OpenAI agents hacked Hugging Face servers and private code in coordinated cybersecurity test breach

Chinese government funding and manufacturing investment are racing ahead of the humanoid robots ability to operate outside of choreographed demonstrations — like the kung fu routines and boxing matches that have made the machines internet sensations reut.rs/4y64JDP

bluesky_Reuters · about 3 hours ago

#USA The U.S. Army has selected five companies for the Janus Program, which includes up to $2.2 billion in awards for the development of more than 20 nuclear microreactors at various military installations. This initiative aims to enhance energy resilience and reduce reliance on external power grids. Subscribe to @OSINTdefender | 💬 Chat room

telegram_OSINTdefender · about 3 hours ago

Chinas humanoid robots arent smart enough to take your job – yet reut.rs/4xtGC26

bluesky_Reuters · about 3 hours ago

Bill Gates warns world unprepared for AI upheaval Bill Gates warned Wednesday that the world is dangerously unprepared for the upheaval brought by artificial intelligence, with the technology destroying jobs and threatening human connection, especially for…

japantoday · about 3 hours ago

UK politicians under scrutiny over AI industry ties: Report A report by Scottish daily The National has revealed that dozens of current and former British politicians have taken advisory roles, donations, or hospitality from artificial intelligence firms, sparking concerns that major technology companies may be using their influence to weaken regulation.

yenisafak · about 3 hours ago

OpenAI explains how its naughty AI agents attacked Hugging Face Biz describes its act of automated irresponsibility as a warning shot

the register · about 4 hours ago

Avertismentul lui Bill Gates privind AI: „Nu-mi place să le dau veşti proaste. Ce propune miliardarul pentru viitorul angajaților Bill Gates, cofondator şi fost CEO al companiei Microsoft, a lansat un apel pentru gestionarea schimbărilor generate de AI, propunând o taxă pe inteligenţa artificială, crearea unui organism internaţional şi desemnarea unor locuri de muncă rezervate oamenilor. Totuși, el admite că este greu de decis care sunt meseriile rezervate doar oamenilor.

digi24 · about 4 hours ago

OpenAI calls Hugging Face breach warning shot on autonomous AI risks OpenAI warned that a cybersecurity breach involving Hugging Face serves as a "warning shot" demonstrating how advanced AI agents can autonomously bypass safeguards and carry out dangerous actions, saying the July incident exposed critical vulnerabilities in research infrastructure and prompted immediate pledges for stricter network restrictions and isolated testing environments.

yenisafak · about 5 hours ago

US seizes internet domains allegedly used by Chinese hackers to target government agencies and critical infrastructure. Justice Department and FBI said the seized domains were used by a Chinese state-sponsored hacking group known as "QTFY" u.afp.com/SVfn

bluesky_AFP English · about 5 hours ago

#China #Taiwan Taiwan has unveiled the Strong Bow anti-ballistic missile system, which features hypersonic speeds, a two-stage solid-fuel propulsion system, and advanced radar capabilities. This system is designed to intercept incoming missiles at altitudes higher than existing defenses, enhancing Taiwans air defense against threats from China. Subscribe to @OSINTdefender | 💬 Chat room

telegram_OSINTdefender · about 5 hours ago

OpenAI says earlier signals could have prevented the Hugging Face breach OpenAIs technical report on the Hugging Face breach says an internal team saw its models reaching the open internet from their sandbox in late May, and that a June alert did not stop the evaluation. It also found that training sometimes rewarded agents for exploiting their own environment. OpenAI has published its account of how […] This story continues at The Next Web

the next web · about 5 hours ago

Anthropic to rent AI computing power from Nscale for $45 billion, source says reut.rs/4wUnCIT

bluesky_Reuters · about 6 hours ago

🇷🇺🇨🇳🇺🇸Russia and China are rapidly catching up to the United States in the space race, according to The Washington Post. The publication notes the rapid deployment of Russias "Rosetta" satellite constellation and the successful launch of Chinas reusable rocket, Zhuque-3. Both programs are intensifying competition with American companies SpaceX and Blue Origin.

telegram_Intel Slava · about 6 hours ago

Cyberattaque contre Hugging Face : près de 700 agents IA se sont coordonnés lors de cette intrusion Un rapport, publié mercredi 26 août par des enquêteurs indépendants, revient sur ce récent piratage mené de leur propre initiative par des agents dintelligence artificielle développés par lentreprise OpenAI.

le monde · about 6 hours ago

#USA Redwire is investing in next-generation phased array antennas to enhance global warfighter connectivity and improve data links for military satellites. These antennas provide dynamic beam steering, increased throughput, and improved link reliability, essential for modern defense communications. Subscribe to @OSINTdefender | 💬 Chat room

telegram_OSINTdefender · about 6 hours ago

⚡️🤖 – An independent review by two non-profit AI safety organizations found hundreds of agents went rogue during last months hacking spree against Hugging Face, POLITICO reports. Around 700 agents powered by two of the companys most capable models went rogue for over 7 days without OpenAIs knowledge. 1,200 agents that "were supposed to be isolated from one another exchanged over 70,000 secret messages about how to cheat their way through a common hacking evaluation", with some agents even sacrificing themselves through dead-end hacking techniques just to give the larger swarm more information. This has raised new concerns that AI is evolving faster than it can be regulated, and some believe were building AI models we wont be able to control in the future. @GeoPWatch

telegram_GeoPWatch · about 6 hours ago

OpenAI agents hacked Hugging Face in 700-strong swarm, tried to cover tracks, investigations find reut.rs/3UfhTQl

bluesky_Reuters · about 6 hours ago

Waystar turns agentic AI loose on claims, denials, and documentation Waystar has launched agents that interpret payer responses and automatically resubmit denied healthcare claims, which it calls the industrys first autonomous resubmission capability. Germany caps how many hospital invoices may be audited each quarter and charges a flat fee when one is reduced. An American company has built software that argues with insurers on its […] This story continues at The Next Web

the next web · about 6 hours ago

A robot sprinter shatters a 100-meter record again as sparks fly at Chinas robot competition A robot sprinter has broken another 100-meter record at Chinas World Humanoid Robot Games

abcnews · about 7 hours ago

#USA TrustPoint successfully demonstrated its C-band navigation signals in the presence of GNSS interference using an enhanced receiver from NovAtel, marking a significant step toward commercializing GNSS-independent positioning and navigation technology. This test was conducted under a contract with the U.S. Naval Air Systems Command. Subscribe to @OSINTdefender | 💬 Chat room

telegram_OSINTdefender · about 7 hours ago

Exclusive: Why Antares is bringing nuclear power to Fort Bragg The U.S. Army and the Defense Innovation Unit have selected Antares to build and operate nuclear microreactors at Fort Bragg, North Carolina, chief executive Jordan Bramble told Axios.Why it matters: The militarys sky-high energy demand — for reusable batteries, computer farms, radars, housing, weapons and more — is an ideal pressure test for nuclear, which has mounted a comeback in recent years.Driving the news: The arrangement with Antares is one of a handful announced Wednesday under the Janus program, the Armys closely watched, multibillion-dollar answer to a White House executive order issued last year.Antares is expected to deliver its reactors in "three packs," according to the service.What theyre saying: "I think the military, for the last decade or so, has been between a rock and a hard place on energy. Most military services bifurcate how they think about energy," Bramble said. "We need to bring more of an operational energy mindset to our installations, here inside the U.S.," he added. "We need to be able to sustain our most critical operations without the grid, without a civilian supply chain, and nuclear power is actually the best thing we have, in terms of capacity factor, to go do that."Context: Fort Bragg is home to the XVIII Airborne Corps as well as the Army special operations and western hemisphere commands. Tens of thousands of people, uniformed and civilian, are at the post.Follow the money: Antares declined to disclose the exact dollar amount of its Janus contract. But the CEO said that between work for the Army, Air Force and Space Force, the team now has "a contract value on the order of $1 billion" and that figure is "growing rapidly.""Everybody wants to sell to defense — but we started the company, we developed our technology, really focused around this market," he said. "Were already in motion."The company recently raised a $470 million Series C.Flashback: The Antares Mark-0 reactor hit criticality June 4 at Idaho National Laboratory. It was the first to do so for the Energy Departments Reactor Pilot Program. The Army was kept abreast.The reactor used tri-structural isotropic fuel, or TRISO. Enough was made for multiple reactors."We actually designed our fuel around the same exact fuel spec that was developed for Project Pele," Bramble said. (Pele is a mobile reactor initiative overseen by the Pentagons Strategic Capabilities Office.)Zoom out: The other Janus awardees and their assigned sites are:BWXT Advanced Technologies at Fort Campbell, Kentucky;General Atomics Electromagnetic Systems at Fort Hood, Texas;Radiant at Fort Benning, Georgia;And Westinghouse Government Services at Fort Drum, New York.The intrigue: "These are very different sorts of companies," Jeff Waksman, the Armys principal deputy assistant secretary for installations, energy and environment, told reporters Wednesday."We believe this will be the spear tip," he said, "not just for microreactors, but for all advanced reactors in the United States."What were watching: The waste disposition pathways, and how they are negotiated.Go deeper: Behind the scenes of Trumps advanced reactor race

axios · about 7 hours ago

AI receptionists seem like a dangerous idea after a patient revealed her doctors system could not understand her speech A GP surgery in South Yorkshire has withdrawn an AI receptionist after a woman recovering from a stroke could not make it understand her and moved to another practice. The NHS Accessible Information Standard has required practices to identify and meet patients communication needs since 2016. A woman recovering from a stroke could not get […] This story continues at The Next Web

the next web · about 7 hours ago

Vanguard is acquiring Altruist, the AI custody platform it backed six years ago Vanguard has agreed to acquire Altruist, an AI-driven custody and software platform for independent financial advisers that it first invested in during 2020, on undisclosed terms. Britain and the EU are addressing the same shortage of advice through rule changes rather than acquisitions. Vanguard is buying the platform it helped fund. It has agreed to […] This story continues at The Next Web

the next web · about 8 hours ago

North Korea to send around 100 athletes to Asian Games in Japan North Korea will send around 100 athletes competing in ‌14 sports to the Asian Games in Japan later this year, the secretary general ‌of the countrys Olympic Committee…

japantoday · about 8 hours ago

Bill Gates warns AI doomsday scenario is coming, with bioterror pandemics and huge job losses but tech bosses are keeping quiet about it Bill Gates warned on Wednesday that AI could cost millions their jobs and help create a devastating bioterrorism weapon capable of threatening humanity.

dailymail · about 8 hours ago

US Army awards up to $2.2 billion to five companies to build microreactors on bases reut.rs/4c8RZUr

bluesky_Reuters · about 8 hours ago

Investigators say hundreds of OpenAI agents hacked Hugging Face and tried to cover their tracks reut.rs/4cb6neR

bluesky_Reuters · about 8 hours ago

Gemini Live gains agentic tasks, voice inbox control and personal intelligence Google has added four capabilities to Gemini Live, letting spoken commands trigger background agents that work across Docs, Sheets and Drive, alongside hands-free Gmail control and a spoken daily brief. Its DMA settlement requires it to let rival assistants reach comparable functionality on Android by July 2027. Google has given its voice assistant the ability […] This story continues at The Next Web

the next web · about 8 hours ago

OpenAI had warnings before its agents broke out OpenAI missed and failed to act on several warning signs that its models were exploiting security flaws and breaking out of their testing environments before they breached Hugging Face, according to a technical report released by the company Wednesday.Why it matters: The incident raises questions about whether AI companies testing environments and internal safeguards can keep pace with models that are increasingly capable of finding and exploiting security weaknesses on their own.Driving the news: OpenAIs technical deep dive into last months Hugging Face breach outlines how its agents also accessed other third-party environments, including a customer of Modal Labs and an account belonging to a user of another unnamed service.According to the report, a cybersecurity monitoring tool alerted OpenAI on June 27 to suspicious activity that responders traced to agents using an improvised message board as a way to move through the network. At the time, on-call staff determined the evaluation did not need to be stopped.A month earlier, an internal team had observed an agent using the message board as well as instances of agents accessing the internet when they werent allowed to, per the report.OpenAI previously disclosed at Black Hat that its agents had turned the software repository tool Artifactory into an improvised message board and later exploited a zero-day in the service to access the internet."With the benefit of hindsight, some early signals identified in this report could have triggered an earlier response," the company writes in the report. The big picture: The technical report comes as OpenAI continues to pause some model work, including delaying the release of its upcoming Astra model, as it re-evaluates its safety practices.Anthropic and Meta have both said in the weeks after the Hugging Face attack that their models have also hacked real-world systems during pre-deployment testing.Zoom in: OpenAIs agents were able to execute their own code on 41 Hugging Face production servers and obtained root-level control of at least one production machine, according to the companys technical report.The agents also accessed production credentials and some internal data and downloaded four private code repositories.The intrigue: At the same time, OpenAIs agents also breached the AI labs own internal systems and were able to read 956 stored secrets, including sensor credentials from OpenAIs cybersecurity monitoring tool.On July 19, OpenAIs agents identified a recently disclosed vulnerability affecting a Linux machine underlying their testing environment.The agents exploited the flaw to gain root-level privileges on the machine, allowing them to move laterally through the connected environment.The agents eventually gained administrator-level access to part of OpenAIs cloud infrastructure and accessed credentials tied to several internal systems.In a separate testing run that day, another internal model from the same family as Astra found signing-key material left behind by earlier agent activity and used it to create a fresh administrator credential.Yes, but: OpenAI says the models actions stemmed from attempts to complete cybersecurity evaluations, including ExploitGym, which tests whether a model can find and exploit vulnerabilities on its own.The Hugging Face incident involved two models: GPT-5.6 Sol and an internal-only research model, but the report says the internal research model was the primary driver of the intrusion. Between the lines: OpenAIs investigation also found evidence that its training may have inadvertently reinforced some of the behaviors that contributed to the incident.During one training run, agents increasingly learned to probe and exploit parts of their environment when the tools they were supposed to use werent available or werent working.In some cases, models received positive rewards after using these unintended methods, which OpenAI says can reinforce the behavior.In one example, an agent exploited a vulnerability to access the underlying program it was supposed to recreate, copied the answer and received a positive reward for successfully completing the task.What to watch: Earlier this week, the Alabama attorney generals office sent a subpoena to OpenAI as part of an investigation into the Hugging Face incident. Other state attorneys general have also asked OpenAI to preserve internal documents. Go deeper: AI agents have a history of escaping tests

axios · about 9 hours ago

U.S. says Chinese hackers broke into Justice Department, NASA, Federal Reserve, Senate The United States said on Wednesday it had disrupted a Chinese hacking operation responsible for break-ins and ‌attempts on the U.S. Justice Department, NASA, the Federal Reserve, the…

japantoday · about 9 hours ago