OpenAI didn’t catch an autonomous breach of another artificial intelligence (AI) company by one of its advanced AI models for a week, and not until after the FBI had been contacted by the hacked company, according to a report.

On Tuesday, OpenAI announced the breach of AI company Hugging Face that happened during one of OpenAI’s internal reviews of several of its models, including GPT-5.6 Sol, calling it an “unprecedented cyber incident.”

“The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities,” the company said. “We are strengthening the containment, monitoring, access controls, and evaluation practices used during model development.”

The hack of Hugging Face started on July 11, and continued until July 13, Thomas Wolf, Hugging Face’s co-founder, told Reuters.

TRUMP LAUNCHES GOLD EAGLE TO HUNT CYBER FLAWS WITH AI

OpenAI revealed on Tuesday one of its AI models had autonomously hacked another company’s infrastructure.  (Omar Marques/SOPA Images/LightRocket via Getty Images, File / Getty Images)

It was several days before OpenAI realized its agent was behind the attack and the two companies didn’t communicate for the first time until July 20, four people, including Wolf, told the outlet.

OpenAI often runs simultaneous model tests, which can make it difficult for employees to monitor everything, four people told Reuters.

Hugging Face told Reuters it is preparing a public timeline of the hack.

According to OpenAI, the incident took place during an internal evaluation designed to measure its AI models’ advanced cyber capabilities. Researchers disabled some built-in safety safeguards and ran the models in an isolated testing environment with limited internet access.

OpenAI said the models exploited an unknown software flaw to access the internet, then breached Hugging Face’s systems in an apparent attempt to find answers to a cybersecurity benchmark.

OPENAI’S SAM ALTMAN WANTS TO NEGOTIATE A 5% STAKE IN COMPANY FOR US IF COMPETITORS AGREE TO KEY PROVISION

Hugging Face said it was preparing a timeline of the hack.  (Jakub Porzycki/NurPhoto via Getty Images, File / Getty Images)

OpenAI said it’s now implementing stricter security controls while vulnerabilities are patched and strengthening safeguards around future AI training and evaluations.

It wasn’t until July 16 after Hugging Face wrote in a blog post that it had been hacked by an “autonomous AI agent system” that OpenAI realized one of its agents was the source, two people told Reuters.

This was a week after the responsible agent first attempted to break out of its OpenAI testing environment.

And by the time OpenAI contacted Hugging Face about the attack, they had already contacted the FBI.

OpenAI told Reuters there were several inaccuracies in its reporting but didn’t respond when asked for specifications.

OpenAI CEO Sam Altman publicly announced the attack on Tuesday. (Sean Gallup/Getty Images, FIle / Getty Images)

OpenAI shared this statement with FOX Business: “We recognize there are a lot of questions and speculative details circulating related to the Hugging Face incident. This is an unprecedented incident, and we think it marks an important moment for AI safety. We are still conducting a thorough review along with external advisors and with oversight from our Safety and Security Committee. Once the review is complete, we plan to publish a technical report of our learnings in the coming weeks.”

The FBI told FOX Business that it declined to comment. 

FOX Business has also reached out to Hugging Face. 

In an X post this week, Hugging Face co-founder and CEO Clem Delangue addressed the incident after OpenAI CEO Sam Altman announced the hack.

OpenAI said one of its AI models compromised another company’s systems during internal testing, prompting a joint investigation with AI startup Hugging Face. (Reuters/Dado Ruvic, File / Reuters)

“We suspected last week’s cyberattack might have come from a frontier lab, given the sophistication of the agent. Turns out it did!” Delangue wrote.

CLICK HERE TO DOWNLOAD THE FOX NEWS APP

He added, “We’ve spent the past 24 hours working closely with the @OpenAI team (thanks!), and we strongly believe there was no malicious intent on their part. It’s quite mind-blowing that all of this happened autonomously! The investigation is ongoing, and we’ll share more learnings from what might be the first incident of its kind!”

FOX Business’ Michael Sinkowitz contributed to this report. 

Investors weighing where to put new money this year have two unusually different but similarly buzzy options on the table: South Korean memory chipmaker SK Hynix, whose U.S.-listed shares debuted with the largest foreign IPO in American history earlier this month, and SpaceX, which completed the largest initial public offering ever recorded just weeks earlier. Both stocks have generated intense investor interest, but their businesses, valuations and risk profiles differ substantially. Here’s what the numbers show.

Note: This article provides factual information to help readers understand each stock; it is not financial advice, and individual investment decisions should account for personal risk tolerance and, where appropriate, guidance from a licensed financial advisor.

How each company reached the public markets

SK Hynix’s American depositary receipts began trading on Nasdaq on July 10, raising $26.5 billion in the largest first-time share sale by a foreign company in U.S. history, surpassing Alibaba’s 2014 offering. Shares priced at $149 and jumped 13% on their debut, and the offering was more than seven times oversubscribed.

SpaceX went public just weeks earlier, on June 12, in an offering that dwarfed even SK Hynix’s. The company sold 555.55 million shares at $135 each, raising more than $75 billion and valuing the company at $1.75 trillion, the largest IPO in history. Shares opened at $150 and surged more than 30% before closing the first day at $160.95, a rally that briefly made Elon Musk the world’s first trillionaire.

Where the stocks trade now

As of Friday, July 24, SpaceX shares, trading under the ticker SPCX, stood at $114.25, down from a previous close of $118.24, and well below the stock’s all-time high of $225.64 reached June 16. That represents a decline of roughly 50% from the stock’s post-IPO peak, according to Yahoo Finance, which also noted the stock is down about 25% over the past month and roughly 9.7% over the past week alone, amid rising short interest and volatility ahead of the company’s upcoming earnings report.

SK Hynix’s ADR, by comparison, has held up considerably better. Shares traded around $172.70 on Friday, up 4.5% on the day, and the stock’s premium over its Seoul-listed common shares has remained elevated, reflecting sustained U.S. investor demand for direct exposure to the company’s memory chip business.

The bull case for each

SK Hynix’s investment case centers on its position as the world’s leading producer of high-bandwidth memory, the specialized chip technology essential for powering the AI accelerators made by companies like Nvidia. The company’s stock has climbed more than sevenfold over the past year amid a global shortage of AI-grade memory chips, and it is scheduled to report second-quarter earnings on July 29, with analyst estimates already trending upward heading into that report.

SpaceX’s bull case rests on a broader, more diversified set of businesses: its dominant position in commercial rocket launches, its rapidly growing Starlink satellite internet service, and long-term optionality tied to Starship, the company’s next-generation reusable rocket system. HSBC, in a note initiating coverage of the stock, modeled a “blue sky” scenario valuing SpaceX as high as $293 per share if Starship becomes commercially viable starting in 2027 and Starlink captures a larger share of the broadband market. Alphabet’s own second-quarter filing disclosed a $94 billion stake in SpaceX, underscoring how deeply other major technology companies have bought into the company’s long-term prospects.

The bear case and valuation concerns

Despite that bullish long-term framing, HSBC set a base-case price target of just $115, below both SpaceX’s $135 IPO price and its recent trading level, and initiated coverage with a hold rating, arguing that shares already reflect much of the company’s long-term growth potential even after applying a premium for Musk’s track record of building disruptive businesses. According to StockAnalysis.com, SpaceX’s roughly $2.77 trillion peak valuation implied a price-to-sales ratio of nearly 150 times its $18.7 billion in 2025 revenue, a figure the site contrasted with defense contractor Raytheon, which trades at roughly 2.8 times its larger revenue base.

SK Hynix, by contrast, has drawn more measured valuation concerns, though its ADR has traded at a substantial premium, at times exceeding 30% to 50%, over its Seoul-listed shares, reflecting strong but potentially overheated U.S. investor demand relative to the underlying Korean stock.

Analyst sentiment

Wall Street’s overall stance on SpaceX remains cautiously optimistic despite the stock’s post-IPO slide. According to Investing.com, 27 of 34 covering analysts rate the stock a buy, versus just one sell rating, with an average 12-month price target of roughly $231 to $237, implying substantial upside from current levels, though estimates range widely from a low of $62 to a high of $800, reflecting deep disagreement about the company’s ultimate trajectory.

Volatility and risk profile

The two stocks differ sharply in recent volatility. SpaceX carries a reported beta coefficient of nearly 6, according to TradingView, reflecting extreme price swings relative to the broader market, and the stock has become the subject of a large short-selling position, with bearish bets reportedly gaining more than $15 billion in value as shares fell from their post-IPO highs. SK Hynix, while itself a historically volatile stock, has shown comparatively steadier post-IPO trading, buoyed by continued strong demand signals in the AI memory chip market heading into its earnings report.

What to weigh before deciding

Investors comparing the two are weighing fundamentally different bets: SK Hynix offers more direct, immediate exposure to current AI infrastructure spending through an established, profitable chip business with a clearer near-term earnings catalyst in its upcoming report. SpaceX offers exposure to a broader, more speculative set of long-duration technologies, rocket launch dominance, satellite internet and future Mars ambitions, with a valuation that several analysts, including HSBC, have described as already pricing in significant future success.

SK Hynix’s July 29 earnings report will offer the next concrete data point for investors assessing that stock’s near-term trajectory, while SpaceX’s own upcoming quarterly results, along with the outcome of its next Starship test flight, are likely to serve as key catalysts determining whether the stock stabilizes after its steep post-IPO decline or continues to face pressure from rising short interest and lockup-related share supply. Both companies remain central to the broader AI and space infrastructure buildout shaping markets this year, but their risk, valuation and volatility profiles differ enough that the right choice is likely to depend heavily on an individual investor’s time horizon and tolerance for the kind of dramatic swings SpaceX shares have already shown since going public.

ChatGPT and its companion services went down worldwide early Saturday morning, leaving users unable to load conversations, send prompts or access saved chat history in what OpenAI later confirmed was its fourth service disruption in as many days.

Outage-tracking service Downdetector said user reports indicating problems with OpenAI began climbing at 5:11 a.m. Eastern time, with the hashtag #OpenaiDown circulating on social media shortly after. More than 3,000 users had flagged the outage on Downdetector by Saturday morning, and “Is ChatGPT down?” began trending on Google search as reports spread.

What went wrong

The outage affected all three of OpenAI’s core services simultaneously: the ChatGPT chatbot itself, the company’s developer-facing API, and Codex, its AI-powered coding assistant. Users encountered a range of symptoms, including sidebar loading animations that never resolved, an inability to send new messages, and error messages citing “too many concurrent requests.” According to reporting from TheNextWeb, users encountered 503 errors carrying the internal label “biscuit_baker_service_me_circuit_open,” a technical signal indicating requests were being blocked before they could reach OpenAI’s servers.

The disruption was global in scope, with reports coming in from users across the United States, Europe, India and Australia, according to Unite.AI, affecting not just the consumer-facing ChatGPT app but also the many outside applications and services that rely quietly on OpenAI’s API infrastructure to function.

OpenAI’s response

OpenAI acknowledged the outage on its official status page shortly after reports began surfacing, stating it was “investigating the issue for the listed services,” naming APIs, ChatGPT and Codex as affected. The company later updated its language, stating it was “experiencing issues” with “elevated error rates” before eventually reporting a mitigation had been applied. “We have applied the mitigation and are monitoring the recovery,” OpenAI said, as the company moved from an “investigating” to a “monitoring” status within roughly an hour of the outage’s onset.

By late Saturday morning, the outage appeared to be resolved. Outage-tracking service StatusGator reported that OpenAI’s status returned to “operational” as of just after noon UTC, noting 554 total user-submitted reports of problems over the preceding 24 hours, with the disruption marked as resolved.

The fourth disruption in four days

Saturday’s outage was notable not just for its scope but for its place in a broader pattern. According to TheNextWeb, the incident marked the fourth service disruption OpenAI had experienced in four consecutive days, a stretch of instability that has drawn increasing attention given how deeply ChatGPT and OpenAI’s underlying infrastructure have become embedded in both consumer routines and third-party business applications.

A history of recurring outages

Saturday’s disruption continues a pattern that has affected OpenAI’s services periodically since ChatGPT’s public launch. The company experienced a roughly four-hour outage on June 4, 2025, followed less than two weeks later by a separate incident on June 17, 2025, that knocked ChatGPT offline for about two and a half hours, with OpenAI attributing that disruption to “an elevated error rate” between 11:20 a.m. and 1:55 p.m. Pacific time. An earlier outage on Jan. 23, 2025, affected users across the U.S., U.K. and beyond for several hours before OpenAI issued a fix. A more severe incident on June 10, 2025, lasted more than 10 hours and also affected OpenAI’s video generation tool, Sora, alongside ChatGPT itself.

Why these outages matter beyond individual users

Because ChatGPT and OpenAI’s API have become foundational infrastructure for a wide range of third-party products, ranging from customer service tools to coding assistants built on top of Codex, disruptions to OpenAI’s core services can ripple outward well beyond individuals simply trying to chat with the AI model directly. Businesses that have built products on top of OpenAI’s API face their own downstream outages whenever the underlying service experiences elevated error rates, a dynamic that has become an increasingly significant point of scrutiny as more companies embed OpenAI’s technology directly into their own offerings.

What affected users experienced

During Saturday’s outage, users described being stuck in persistent loading loops when attempting to open the ChatGPT sidebar or access prior conversation history, according to BleepingComputer. Attempts to send new messages returned errors citing an excessive volume of concurrent requests, effectively locking users out of active conversations even when the interface itself appeared to load normally.

How OpenAI communicates during outages

OpenAI maintains a public status page that the company updates during service disruptions, though as with several past incidents, the company’s public acknowledgment lagged behind the volume of user reports appearing on independent tracking platforms like Downdetector. That pattern, users noticing and reporting problems before OpenAI’s own status page reflects an issue, has remained consistent across the company’s outage history, from January 2025 through Saturday’s incident.

With OpenAI reporting the outage as resolved by late Saturday morning, the company has not disclosed a specific root cause for the disruption or detailed what changes, if any, it plans to make following its fourth outage in four consecutive days. Given the frequency of the recent disruptions, users and businesses relying on OpenAI’s services may reasonably expect continued scrutiny of the company’s infrastructure reliability in the days ahead, particularly as ChatGPT, Codex and OpenAI’s API continue to see growing adoption across both individual and enterprise use cases. For now, service has been restored, but the string of back-to-back incidents this week leaves open the question of whether Saturday’s fix will hold or whether additional disruptions may follow.

The Mega Millions jackpot climbed to $743 million ahead of Friday night’s drawing, with a cash value of $323.4 million, making it the 10th-largest prize in the game’s history as players across the country checked their tickets against Friday’s winning numbers.

The winning numbers drawn Friday, July 24, were 2, 5, 42, 44, 60 and a gold Mega Ball of 1. Drawings are held twice a week, every Tuesday and Friday at approximately 11 p.m. Eastern time.

A long stretch without a jackpot winner

The last time a player won the Mega Millions jackpot outright was more than three months earlier, when a ticket purchased in Ohio claimed a $60 million prize on St. Patrick’s Day, March 17. Since then, the jackpot has climbed steadily through repeated rollovers as no ticket has matched all six numbers in the twice-weekly drawings.

That stretch has not been entirely without winners, however. In Tuesday’s drawing preceding Friday’s jackpot, no ticket matched all six numbers for the grand prize, but two players came close to the top prize: one ticket holder in California won $2 million, and another in Arizona won $3 million, both by matching enough numbers to claim a lower-tier prize under the game’s payout structure.

How the numbers compare historically

Friday’s $743 million prize now ranks 10th on the all-time list of largest Mega Millions jackpots, a list still topped by the record $1.602 billion jackpot won in Florida on Aug. 8, 2023. The full top 10 includes $1.537 billion won in South Carolina on Oct. 23, 2018; $1.348 billion won in Maine on Jan. 1, 2023; $1.337 billion won in Illinois on July 29, 2022; $1.269 billion won in California on Dec. 27, 2024; $1.128 billion won in New Jersey on March 26, 2024; $1.05 billion won in Michigan on Jan. 22, 2021; $983 million won in Georgia on Nov. 14, 2025; $810 million won in Texas on Sept. 10, 2024; and Friday’s $743 million prize, with a winner still to be determined.

What winners can choose

Should a single ticket match all six numbers, the winner would face a choice between two payout options: a one-time lump-sum cash payment or an annuitized prize paid out through an initial payment followed by 29 annual installments, each 5% larger than the one before it. Most jackpot winners historically opt for the lump-sum cash payment rather than the long-term annuity structure.

How the game works

Mega Millions is played across 45 states, the District of Columbia and the U.S. Virgin Islands. Each ticket costs $5, and players select six numbers across two separate pools: five different numbers from 1 to 70, known as the white balls, and one number from 1 to 24, known as the gold Mega Ball. Players uncertain of their preferred numbers can opt for a “Quick Pick” or “Easy Pick” selection, which lets a terminal generate a random set of numbers on their behalf.

To win the jackpot, a ticket must match all six numbers drawn. In the rare event that more than one ticket matches all six numbers in a single drawing, the jackpot prize is split evenly among the winning tickets.

Mega Millions tickets now come with a built-in multiplier that automatically boosts non-jackpot prize winnings by two, three, four, five or 10 times. That feature replaced the previous system, under which players had to pay an additional dollar to add the optional “Megaplier” to their ticket.

How long winners have to claim a prize

According to Mega Millions, the window for collecting winnings varies by state, typically ranging from 90 days up to a full year from the date of the drawing. Lottery officials encourage players to check with their specific state lottery for the exact claim period that applies to their ticket, since deadlines and procedures are not uniform nationwide.

Where to buy tickets

Mega Millions tickets can be purchased at a range of locations, including convenience stores, gas stations and grocery stores. In some states, tickets are also available for purchase online, offering an additional option for players who prefer not to buy in person.

A broader trend in jackpot growth

Friday’s near-record prize continues a broader pattern of increasingly large Mega Millions jackpots in recent years, driven in part by a 2017 overhaul of the game’s number pool that lengthened the odds of winning the jackpot from roughly 1 in 259 million to about 1 in 302.6 million. Longer odds have generally meant more frequent rollovers and, correspondingly, larger jackpots by the time a winning ticket is eventually sold. Rising ticket prices, which moved from $1 to $2 following the 2017 changes and then to the current $5 price point more recently, have also factored into the size of jackpots the game has produced.

Separately, Powerball, Mega Millions’ chief rival among major U.S. lottery games, recently announced a structural change of its own that officials say could allow its jackpots to grow at a faster pace going forward, a shift that could intensify competition between the two games for players chasing the next big prize.

The next Mega Millions drawing is scheduled for Tuesday, July 28, at 11 p.m. Eastern time. Should no ticket match all six numbers in Friday’s drawing, the jackpot will continue to grow heading into that next drawing, potentially climbing further up the list of the game’s largest-ever prizes. Players hoping to check their Friday tickets against the winning numbers, 2, 5, 42, 44, 60 and Mega Ball 1, are encouraged to verify results directly through their state lottery or the official Mega Millions website.