Compounded GLP-1 injections made by a pharmacy previously warned by the Food and Drug Administration about putting “patients at risk” have been recalled nationwide.
Apothecary Pharma LLC first initiated a voluntary recall of tirzepatide injections, the active ingredient in Zepbound and Mounjaro, on Sept. 24 for “lack of assurance of sterility,” according to an FDA notice. The federal agency updated the recall on Oct. 10 to classify it as a “Class II” recall, meaning using the medication could cause “temporary or medically reversible adverse health consequences.” However, the risk of serious harm is small.
The injections are compounded versions of the GLP-1, meaning they were not name-brand or FDA-approved and were custom formulated by a specialized pharmacy called a compounding pharmacy.
Sterility and sanitation issues have stirred controversy around compounding pharmacy injections in the past, such as a 2012 case in which nearly 800 people became ill and 64 died from a fungal meningitis outbreak caused by unsanitary injections from the New England Compounding Center.
Compounded GLP-1s have become a special focus of the FDA since the removal of tirzepatide and semaglutide from the official drug shortage list in late 2024 and early 2025, meaning compounding pharmacies that filled the gap for millions were expected to stop producing them en masse. Many have continued, prompting the agency to issue warnings and various legal challenges.
In December 2025, Apothecary Pharma received an FDA warning letter listing several violations, including “serious deficiencies in your practices for producing drug products intended or expected to be sterile, which put patients at risk.”
USA TODAY contacted Apothecary Pharma and the FDA for comment on Oct. 19, but has not received a response.
Which compounded GLP-1 injections were recalled?
A total of 22 vials of tirzepatide were recalled nationwide.
Apothecary Pharma Tirzepatide Injection: 18mg/mL, 4 mL sterile multi-dose vial, Rx only, with the following lot codes and beyond-use dates (BUD):
Lot AP-202622, BUD 07-Oct-2026
Lot AP-202623, BUD 08-Oct-2026
Lot AP-202624, BUD 19-Oct-2026
Lot AP-202625, BUD 10/20/2026
Lot AP-202627, BUD 10/26/2026
Lot AP-202628, BUD 10/27/2026
Lot AP-202629, BUD 10/28/2026
Lot AP-202630, BUD 10/31/2026
Lot AP-202631, BUD 11/01/2026
Lot AP-202632, BUD 11/03/2026
Lot AP-202633, BUD 11/07/2026
Lot AP-202634, BUD 11/08/2026
Lot AP-202635, BUD 11/16/2026
Lot AP-202636, BUD 11/17/2026
Lot AP-202637,BUD 11/18/2026
Lot AP-202638, BUD 11/19/2026
Lot AP-202639, BUD 11/20/2026
Lot AP-202640, BUD 11/21/2026
Lot AP-202641, BUD 11/22/2026
Lot AP-202642, BUD 11/24/2026
Lot AP-202643, BUD 11/25/2026
Lot AP-202644, BUD 11/26/2026
What to do with recalled GLP-1 injections
The FDA and Apothecary have not issued a public statement on what to do with affected vials; meanwhile, customers online have reported requests through their prescribers to return or dispose of them.
If you have an affected lot, do not use it. Contact your prescriber for instructions on what to do with recalled medication.
What are compounding pharmacies? Are they safe?
Compounding pharmacies, or special pharmacies that create customized versions of drugs from raw ingredients, have sparked renewed interest in recent years thanks to the surge of GLP-1s and the peptides industry.
Unlike a commercial pharmacy, which sells mass-produced drugs, a compounding pharmacy can develop specific medications to meet a patient’s needs, including by adjusting the dose or removing allergens like dyes. Compounding pharmacies are often used for pediatric and veterinary cases where a medicine needs a different format, such as a lower-concentration liquid for a small animal.
These pharmacies are also sometimes used during drug shortages, as the FDA grants temporary approval for them to produce a drug in bulk to fill the void.
This was the case in 2022 when the FDA gave compounders the go-ahead to begin producing their own versions of GLP-1 medications amid a shortage. Tirzepatide and semaglutide were removed from the official drug shortage list in late 2024 and early 2025, but compounding pharmacies have continued to produce their own versions and then sell them to millions, especially through telehealth platforms like His/Hers.
Compounding pharmacies must use FDA-approved ingredients, but the actual formulations themselves aren’t evaluated for purity, safety or effectiveness, including mass-produced medications. Compounding pharmacies also aren’t routinely FDA-regulated, which has led to sanitation and contamination issues, such as the 2021 New England Compounding Center case or 2023 FDA reports of compounders using unapproved salt forms of ingredients in their formulations.
These factors have become increasingly controversial, with pharmaceutical giants like Eli Lilly and Novo Nordisk suing compounders to stop or limit the practice.

On Monday, Amazon Web Services released a tool called TorchNeuron that lets AI startups easily move their AI apps onto Amazon’s AI chip, Trainium. In doing so, Amazon is also taking a direct stab at the the software advantage that has helped Nvidia become the default hardware choice for AI apps, and spurred its rise as AI hardware provider of choice.
TorchNeuron works with PyTorch, a free, open-source toolkit that developers use to build, train, and run AI models. It lets AI apps written with PyTorch run on Trainium by changing just one line of configuration, according to Amazon — no code rewrite required.
Amazon says about 85% of AI apps are written with PyTorch. Facebook originally developed it, but since 2022, it has been part of the Linux Foundation, which maintains open-source projects. Nvidia has spent years ensuring its CUDA software — the software infrastructure that allows apps to run on Nvidia GPUs and its other hardware — seamlessly supports AI apps written with PyTorch. As a result, Nvidia’s hardware is now deeply embedded in the toolkit, essentially by default.
TorchNeuron is adapted from PyTorch, so developers who know PyTorch already know how to use it. Amazon isn’t the first company to attempt to easily bring their chips to the platform. PyTorch users can also port their apps to AMD GPUs or even Apple’s chips with a simple configuration change.
However, Amazon argues that porting to Trainium isn’t just easy; it also gives AI startups access to Amazon’s entire cloud. Unsurprisingly, AWS claims that this is a better choice for startups than using neoclouds, many of which are tied to Nvidia and its hardware, like Coreweave, Nebius, and Firmus.
“You’re building on AWS. You’re not betting on a single-purpose compute provider and scrambling to build everything else around it,” Colin Brace, VP of Technology Software at AWS Annapurna Labs (the unit that makes Amazon’s custom chips) told TechCrunch via email. “On AWS,” he continued, “the infrastructure to go from prototype to production is already there, allowing a founder to move fast and scale when they need to.”
AWS says that because the same chip both handles training (teaching the AI model) and inference (putting the finished model to work, as when it answers a user’s question), and is tightly integrated with the cloud infrastructure that startups need, its cloud will save them money, even if AWS doesn’t offer the cheapest per-token price.
“Trainium’s value is price-performance plus the fact that it’s one platform for both training and inference, which improves the economics across a full workload,” Brace said. “Look at Deepgram — they moved to Trainium using native PyTorch, support and they’re processing 70% more audio per dollar at production-grade accuracy.” Deepgram is an AI voice app provider, a competitor to ElevenLabs.
But Amazon isn’t content to just give PyTorch app developers an easy option to move to its cloud.
Amazon also says it is pushing PyTorch to be less automatically centered on Nvidia, and it will contribute the code to do that to the main PyTorch project. In the blog post announcing the new tool, the company wrote:
“Amazon’s approach extends beyond its own hardware. The company is contributing changes upstream to PyTorch itself, working with the PyTorch community to remove code that ties the framework to specific chips. The goal is to make PyTorch more hardware-agnostic for the entire developer community, not just Trainium users.”
While such a tactic is undoubtedly good for those who are selling hardware and cloud services that compete with Nvidia, one of the points of PyTorch is to hide all the pesky difficulties of getting AI code to run on hardware. We’ll have to wait and see if pushing Nvidia out that far gives developers more choice and more affordable options, as Amazon promises, or just more configuration headaches.
Either way, if Amazon succeeds in deeply uncoupling PyTorch from Nvidia, this could have much bigger ramifications for the GPU giant than simply dealing with chip competitors. As Brace said: “The real story is removing the lock-in that kept developers from having a real choice at all.”

A former OpenAI engineer who allegedly installed a sophisticated listening device in an executive conference room was in close contact with a member of the Saudi royal family and was planning to leave the U.S. for Riyadh when authorities arrested him, federal prosecutors said Sunday in a court memorandum.
Abdulrahman Alfozan is accused of tapping artificial intelligence tools to help him develop encrypted recording devices and identify meeting rooms used by OpenAI executives, government officials and business partners.
The Saudi national earned an annual paycheck of $385,000 at OpenAI and received a $2.2 million equity grant when he first joined the company in July 2025.
In an 11-page filing, prosecutors wrote that investigators recovered evidence suggesting Alfozan may have planned to place even more devices inside OpenAI offices. Prosecutors also revealed that the suspect engaged in WhatsApp messages with an unidentified Saudi royal who repeatedly encouraged him to return to the kingdom and join a Saudi AI initiative.
The filing also shows photographs of other potential surveillance locations, internal searches concerning meetings between OpenAI and Saudi government officials, and notes suggesting he planned to take back the devices before leaving the company.
Investigators allege that on Sept. 28 Alfozan installed a “surreptitious audio recording device into a USB port on the back of a television in an executive OpenAI conference room, where the device remained for at least multiple days.”
Prosecutors allege that he “modified the device to prevent data transfer between it and the television into which it was inserted” and “had also encrypted the content stored on the device’s microSD.”
In the days before the device was installed, investigators claim that Alfozan used an AI chatbot account belonging to his wife to research how to customize the recording device, including how to: “build a continuous audio recorder,” “captur[e] microphone audio and save compressed WAV recordings to an SD card,” “improve audio and live listening,” “develop … speech-triggered recording,” “add browser access and controls” including “live listening,” and “encrypt the SD card” and “decrypt and recover recordings from the removed card.”
Meetings in that executive conference room, according to prosecutors, “frequently involve discussions of highly confidential technical, commercial, and strategic information, including proprietary research, as well as nonpublic information about OpenAI’s technical progress and strategic decisions, capabilities, priorities, government engagements, and commercial relationships.”
According to the filing, investigators determined that in the “days surrounding the time period that the device was installed, OpenAI executives met with government officials in that very room,” alleging the “defendant’s conduct could pose a danger to the national security of the United States.”
Prosecutors also disclosed that his cellphone contained photographs of at least one other concealed recorder as well as electronic equipment in other OpenAI meeting rooms.
Prosecutors allege that throughout 2025 and 2026, Alfozan “was in frequent communication via WhatsApp with a member of the Saudi Royal Family, including as recently as October 5, 2026.”
The Saudi royal, according to prosecutors, told Alfozan that an artificial intelligence initiative was being launched in Saudi Arabia and enticed him to join it.
“Saudi needs you more than [sic] the US!” the member wrote, part of a campaign to recruit the defendant. Alfozan agreed but said he wouldn’t leave until his stock vested.
The suspect was arrested Oct. 9, three days before he was scheduled to depart San Francisco with his wife and infant son on a one-way flight to Riyadh, Saudi Arabia.
Instead, his wife and 3-month-old son left the country the following day on business-class tickets purchased by a Saudi military official, prosecutors allege. Before his arrest, the defendant had been scheduled to take two months of parental leave beginning in October 2026.
Alfozan, who has no prior criminal record, according to the filing, became a U.S. permanent resident in September 2020 and applied for citizenship thereafter. OpenAI fired him after his arrest. A detention hearing was scheduled for Monday, before U.S. Magistrate Judge Alex Tse.
The government has argued that Alfozan should be detained pending his trial, alleging that his wealth, connections and apparent assistance from Saudi government officials put him at risk of fleeing the country.

United Airlines is adding a few new destinations to its friendly skies roster.
The airline announced this week that it will add five new nonstop North American routes, with three serving cities north of the border.
Of the new routes, three will be international expansions, and two will be domestic.
Experts say the domestic routes between Houston, TX and Joplin, MO, and between Washington Dulles and Lancaster, PA, stem from ongoing construction and subsequent operational restrictions at Chicago O’Hare International.
Meanwhile, the international additions of Denver to Ottawa, Denver to Quebec City, and Newark to St. John’s reflect the recent transborder business agreement between United and Air Canada.
Some of these Canadian routes are a welcome return while others mark the establishment of a completely new connection.
The Denver-to-Ottawa service re-establishes a former United Airlines route discontinued in 2010, alongside a fresh link between Denver and Quebec City.
Additionally, United will return to Newfoundland and Labrador for the first time since 2016 with seasonal flights connecting Newark and St. John’s.
Domestic service will begin this spring while the Canadian connections take flight this summer.
Meanwhile, to complement these additions, Air Canada is offering new connections between Toronto and Louisville and Quebec and Indianapolis, with seasonal service expected to run June through September of next year.
Air Canada is also taking off with new flights between Edmonton and Newark.
According to Bloomberg, United and Air Canada are looking to add as many as 10 additional US-Canada itineraries
“Air Canada and United’s long-standing partnership continues to deliver meaningful results for our customers and communities. For nearly three decades, Air Canada and United have worked to facilitate seamless travel and strengthen connectivity between Canada and the United States,” said Mark Galardo, Executive Vice President & Chief Commercial Officer and President, Cargo at Air Canada, on October 16.
The added routes come amid a slew of other switches United Airlines has made of late.
Earlier this month, United launched a cheaper first-class fare.
Back in February, United quietly updated a noise policy that — before the change — had flyers nearly losing their minds.
The 100-year-old company told The Post it has started forcing passengers to wear headphones when listening to content on phones, tablets or anything that makes sound.
United’s Contract of Carriage — a rules and policy handbook — was updated on February 27 to add the new “headphone language.”

Kids will soon have a financial interest in the future of space.
“Next year, I will give one of my SpaceX (SPCX) shares to every kid in America. This will be the first of many,” SpaceX founder and CEO Elon Musk said in an X post on Monday.
Musk didn’t specify an age range for the proposed gift, but a Bloomberg report indicated he could use Trump Accounts as the vehicle for his stock donations.
SpaceX stock rose 1% in morning trading on Monday.
There are currently about 73 million children under 18 living in the US, according to the latest Census Bureau data. That number goes down to about 18.5 million for newborns up to 4 years old.
SpaceX COO Gwynne Shotwell has already pledged a personal donation of more than $320 million in SpaceX stock to fund Trump Accounts for more than 2 million children.
Brian Sozzi is Yahoo Finance’s Executive Editor, host of the Sozzi Unleashed morning show and the Power Players with Brian Sozzi podcast, and a member of Yahoo Finance’s editorial leadership team. Follow Sozzi on X @BrianSozzi, Instagram, and LinkedIn. Tips on stories? Email brian.sozzi@yahoofinance.com.
Click here for in-depth analysis of the latest stock market news and events moving stock prices