If youve already given up on your 2026 rebrand because you couldnt stick to your six gym sessions a week and no-sweet-treats resolutions, adopting a vegan plus bacon mindset may be the answer to all your problems.
TikTok creator @addietheoptimist broke the idea down in a recent video: Someone on here went viral because they said if you think you cant go vegan because you love bacon too much, just become vegan plus bacon, she explained in the now-viral clip.
Im here to tell you you can just apply that mentality to so many things in your life.
While the original creator was referencing harm reduction in relation to veganism (that if you only eat bacon sometimes but are 100% vegan otherwise, its still making a difference), the potential applications of the general concept behind vegan plus bacon are limitless. Dont feel like going to the gym? Rather than not going at all, go and give 10%… eve if that looks like 10 minutes walking on the treadmill, scrolling TikTok.
You dont have to be imprisoned by your own rules, the creator concluded.
The video currently has over two million views, with the comments full of examples of instances where others have unknowingly adopted the vegan plus bacon mentality.
One person quit smoking, but still permits the occasional cigarette while among friends. Another is pescatarian, but allows themselves steak once or twice a year. Others are California sober, when a person gives up on alcohol and hard drugs, but continues to smoke weed.
Perfect is the enemy of done, as one comment read. Anything worth doing 100% is still worth doing at 10%, another suggested.
Perfectionism allows no room for mistakes or occasional slipups. Something has to be perfect, or its not good enough. Whats more, researchers found rates of perfectionism have surged in recent decades, having been a core part of Western culture since the 80s. There could be a number of causes for this, including pressure from school or parentsand today, social media also creates an additional pressure to portray a perfect image online.
But perfectionistic tendencies have been linked to an alarming number of clinical issues, including depression and anxiety, even in children, as well as early mortality and suicide. Meanwhile, one of the most effective protections against anxiety and depression is showing compassion to oneself.
Here, the vegan plus bacon mindset acts as the perfect counter-narrative to a culture unhealthily obsessed with self-improvement. Sometimes good enough is good enough.
And, as I recently read somewhere (i.e., saw on TikTok): such a mindset shift only has to make sense to me, and I dont feel like explaining it to anyone else.
A few weeks ago, my iPhone woke me up at 3:30am.
In a way, it was my fault. Id technically set the alarm for three thirty. In another way, it was Siris fault. Apples AI assistant never thought to question if I was asking for AM or PM. And it simply assumed the most painful option for me.
Its just one of countless tiny examples of how Siri, 16 years since Apple acquired the technology, has been a disappointing product. Siri was already looking dusty before modern LLMs, and with the launch of ChatGPT, it has been completely left behind. Which is why in June 2024, with fans and investors growing impatient, Apple promised a new era of AIApple Intelligence.
But that new intelligence never arrived. Reports of infighting at Apple abound, talent left the company last year as Tim Cook held an emergency all-hands in August promising a bigger focus on AI.
This week we learned the results of that renewed focus. Instead of building a new Siri itself, Apple will license AI from Google, and stick Gemini inside its assistant. Siri will get this more personalized, capable upgrade later this year.
This is a significant announcement. Apple had toyed with third party AI before, most recently via an agreement with OpenAI where users could opt in to have Siri ask ChatGPT a question. But this approach was always a silly stopgap. No one wants a go-between between an incapable chatbot and an all-knowing AI. People remove friction from their workflows. They want to just talk to the AI that knows the stuff.
The agreement with Google eliminates middle men. Specifically, it doesnt just mean Siri might send a question over to Gemini to get an answer. It means that Siri is Gemini. Siris foundational architecture is Googles AI tech.
Apple is Google now.
[Illustration: FC]
Is this a failure for Apple? (No)
Its easy to look at this moment as cementing Apples own failure, as the company underinvested in artificial intelligence for years. But while Im hardly an Apple apologist, Id argue exactly the opposite.
Apple doesnt need to make every part of its products to make those products successful.
Yes, Apple has been designing its own mobile processors, for instance, since 2010. These processors have allowed Apple to push the boundaries of performance while cutting back power consumption. They are a big reason that Apple phones and laptops are so appealing to creative professionals.
But Apple doesnt use manufacturing partners build its own screens, which it buys from LG, Samsung, and BOE Technology. It doesnt make its own RAM, either, which is why so many Apple execs are currently holed up in South Korea, hoping to secure the vital hardware from Samsung as its recently ballooned in price.
Even in this challenging supply chain, Apple simply doesnt need to build anything that it cant build best. Screens and RAM are commodities that Apple can acquire, integrate into their products, and no one will know the difference.
Increasingly, AI is a commodity, too.
Thats one reason that, for Apple, this deal with Google is a steal. While headlines focus on the shocking nature of Google powering Siri, its a fact that, I suspect, most iOS users will forget about in day-to-day use as they encounter more and more touchpoints of Apple Intelligence.
Consider that Google pays Apple more than $20 billion a year to be the default search engine on Safari. Cleary, thats considered a worthwhile investment for Google, to get its search engine front and center on Americas most popular phone brand. But all Apple has to do for those dollars is set up a URL. Theres no significant technological investment on Apples part.
If Apple was training an advanced LLM on its own, it would cost a lot of money now, and even more into the future.
Thats because, on one hand, yes, AI models are increasingly commoditized, and the core science driving them is more widely understood than it was a few years ago. On the other, developing those modelsand specifically, staying on the cutting edge in an industry that feels new every three monthsis still a costly practice thats growing costlier by the day.
According to Bloomberg, Apple is paying Google $1 billion a year for AI. Thats a figure remarkably close to the increasing cost of developing AI. Whereas the ChatGPT 3 of 2020 cost a mere $2 to 4 million to train, the latest generations of LLMs built upon vastly more data (like ChatGPT 4 and the latest Gemini) have cost over $100 million to train apiece, with the vast majority of cost going to computing power. And some rumors point to OpenAIs ChatGPT 5 costing another jump more than what weve seen spent thus farbetween $500 million and $2.5 billion according to analysts.
Meanwhile, Apple is getting the worlds top AI model for a mere 1% of its annual profits. $1 billion is barely a line item for a company with a company boasting approximately $400 billion in annual revenue.
If Apple has done nothing else than lock in a $1 billion rate for the near future, to have a top tier AI service from a proven competitor thats paying 20x that to you for a search referralthen thats a considerable business win. Let the engineers focus on tuning an AI to Apples specifications. Then Apple can focus less on building a competent AI and more on how that AI appears across its products.
Indeed, if we should question anything, its how much Apple can lead the industry, not through AI models, but through the design and manifestation of the AI experience.
What does the Apple mean in Apple Intelligence? Now that its sitting on a solid technological foundation, thats the question Cupertino finally gets to answer in 2026.
Get ready with me as an absolute piece-of-shit ICE agent, one such video begins, posted by comedian Adam Macias. These videos are comedic skits, rather than from the social media accounts of actual ICE agents, but have quickly racked up hundreds of thousands of views as the US Immigration and Customs Enforcement agency faces intense backlash after the shooting of Renee Good in Minneapolis.
I wake up screaming and shooting. Not because Im scared but because my sleep paralysis demon looks like someones tía, the skit begins (tia is Spanish for aunt). I start the day off with a shower but no matter how hard I try I just cant seem to get clean, it continues.
After doing laundry and eating breakfast, the ICE agent proceeds to have his second shower of the day. Time for some vocal warm ups, he then says. Self defense, self defense, he yells repeatedly whilst staring himself down in the mirror.
Time to go to work, he continues. My commute isnt far because Im literally about to go deport my neighbor.
On Wednesday, January 7, ICE agent Jonathan Ross fatally shot Good in her car as she was protesting against the agency’s operations in the Twin Cities. The killing sparked widespread backlash and protests, both on and offline, with thousands in attendance at the ICE out of Minnesota rally and march over the weekend.
Wow what a lovely, beautiful day to harass people who are at work, another TikTok creator quips, in a video titled How ICE agents wake up in the morning, viewed almost 3 million times.
I hope I get to beat up a pregnant woman, he says, referencing a recent incident where ICE agents detained a woman who was allegedly pregnant, kneeling on her and dragging her through the snow.
Maybe today Ill get a DUI with my children in the backseat, he added, again referencing a real incident where an ICE agent was arrested for allegedly driving drunk while he had two young children in his vehicle.
Nobody cared who I was until I put on the mask, he concluded the video (a reference from The Dark Knight Rises). Now they just hate me.
Since July, there have been 13 recorded instances of ICE agents firing at or into civilian vehicles, a recent investigation by the Wall Street Journal found. In eight of those instances, civilians were shot, and two confirmed dead. At least five were U.S. citizens.
Whether its in-person demonstrations or posting skits online, Americans arent afraid to make their feelings known. As one comment read: This is actually an official ICE training video.
A political showdown over Greenland, and possibly a military one, is looking increasingly possible as President Donald Trump doubles down on his long-standing threat to take over Denmark’s semiautonomous territory “one way or the other.”
In the wake of the U.S. invasion of Venezuela, Europe’s major military powers are taking Trump’s ramped-up rhetoric with renewed seriousness. And in what appeared to be unthinkable just months ago, the escalating geopolitical conflict could pit the United States against its longtime NATO partnersa troubling sign of the U.S.’s eroding relationships with its closest allies.
Here’s what to know about the situation.
Denmark, Greenland, and U.S. to meet at White House
The U.S. and Denmark, both members of the NATO military alliance, are scheduled, along with Greenland’s foreign minister, to meet on Wednesday at the White House. Both Greenland and Denmark have made it clear that Greenland is not for sale, and their people do not want to be a part of the United States.
Europe, NATO respond to U.S. threat
Greenland’s Prime Minister Jens-Frederik Nielsen and Danish Prime Minister Mette Frederiksen made their position known in a joint press conference on Tuesday: If we have to choose between the USA and Denmark here and now, we choose Denmark. We choose NATO . . . and the EU, Nielsen said, according to Danish public broadcaster DR.
A day earlier, Andrius Kubilius, the European commissioner for defense and space, said that if Denmark requested help in the event of U.S. military action, the European Union could provide security for Greenland, which he said would, in effect, end the NATO alliance, per Reuters.
Also on Monday, NATO Secretary General Mark Rutte said, “When it comes to the Arctic region . . . collectively as an alliance, the collective defense [of Greenland] is crucial . . . and we do everything to protect that territory.”
Why does Trump want Greenland, anyway?
Trump, for his part, has said taking over Greenland would prevent Russia and China from grabbing the strategically located mineral-rich region in the Arcticwhich has a wealth of natural resources and sits on the shortest route between the U.S. and Europe.
Beijing, however, has argued this explanation is merely “an excuse” for Trump to justify more territorial expansion of the U.S.
Remind me, how did we get here?
Back in March, the Trump administration sent an uninvited delegation to the autonomous Danish territory, which, at the time, Greenlands Prime Minister Mute Egede called highly aggressive and a provocation. That trip included second lady Usha Vance, national security adviser Mike Waltz, and Energy Secretary Chris Wright.
Back in November, Fast Company and Johns Hopkins partnered for the first-ever World Changing Ideas Summit in Washington, DC, an event that convened leaders across business and academia to engage with the ideas and innovations reshaping the future.
Knowing we were heading into a new year that undoubtedly is bringing new challenges to every industry, we asked some of our speakers working in space, healthcare, AI, and the intersections therein, what would be top of mind for them in 2026:
Were in a race against resistance.
Akhila Kosaraju, founder and CEO of Pharebio, is using predictive and generative AI to power drug discovery. The startup plans to develop 15 new antibiotics by 2030 that will outpace microbial resistance and cure common but difficult to treat illnesses such as E. coli and C. diff. For Kosaraju, the biggest challenge is time.
Were in a race against resistance, she says, noting it can take fewer than six years for bacteria to develop resistance to antibiotics. Meanwhile, it can take 13 yearsmore than double that timeto develop new medicine.
Phare Bio President and CEO Akhila Kosaraju [Photo: Sardari Group for Johns Hopkins University]
The biggest challenge for us is getting these antibiotics into preclinical and then ultimately clinical development, and FDA approved. I think you’ll hear this across AI for drug discoverythere is an impatience to make sure that these AI models are actually working and that patients actually start seeing those results,” she says. “So our team is laser focused on not just building the best model for its own sake, but really ensuring that we are truly accelerating timelines to get these drugs developed, that these drugs are more novel, less toxic, and more efficacious than what currently exists.
If we don’t have a solution, America’s presence in space will stop.
Saleem Miyan, CEO of Max Space, a company developing new space stations in partnership with the U.S. government and NASA, believes its essential for both American innovation and security to expand the nations presence in spaceespecially given that the International Space Station will be decommissioned in 2030.
That will mean, if we don’t have a solution, America’s presence in space will stopand we don’t want that to happen, Miyan says. As a result of that, we’re putting an awful lot of time, energy, effort, [and] investors capital, to make sure that people understand quite how important being in space is, remaining in space is.
Max Space CEO Saleem Miyan [Photo: Sardari Group for Johns Hopkins University]
And theres no shortage of opportunities for Max Space to capitalize on, Miyan says.
We’re seeing so much attention right now from all governments around the world to get the next space race won, and we’re doing everything we can to make sure that America maintains that position,” he says. “I think in order to do that, we have to make sure that we are continuing to evolve these incredible ideas that companies like ours have developed from the concept all the way through to productization, so that we as a nation maintain this strength and dominance that we’ve always hadnot just in exploration but also in defense.
space is open for innovation.
Were trying to get the message out that space is open for innovation, says Molly Mulligan, Chief Operating Officer of SpaceMD, a subsidiary of Redwire Space. The company, which formed in August 2025, is focused on accelerating drug creation by removing gravity from the growth process of antibody crystals.
“At Space MD, that’s really the goal: to take the science of making better drugs in space and apply that here on Earth to actually make drugs that are commercially available for people, Mulligan says.
SpaceMD Chief Operating Officer Molly Mulligan [Photo: Sardari Group for Johns Hopkins University]
SpaceMDs mission takes inspiration from the success of Keytruda, an immunotherapy drug created by pharmaceutical company Merck whose crystals were grown on the International Space Stationa process they are hoping to replicate in the coming years.
We want to be the company that takes that next drug through the whole process. And we’re working on that today, trying to take a drug from growing the crystals in space all the way through the preclinical process to an investigational new drug form for the FDA,” she says. “So that’s really top of mind for us is being the company that does that with the second drug and the third drug and a fourth drug to help people every day here on earth.”
“patients are waiting for new medicines today.”
Isomorphic Labs, the AI-powered drug company spun out of Google DeepMind, has an audacious goal of solving all diseases using AI. The key to doing so largely rets with its AlphaFold AI system which can predict the complex structure of proteins and model how they might internet with, say, DNA or drugs. The discovery earned Google DeepMind the 2024 Nobel Prize in chemistry. And this past July, the company announced it’s “very close” to human trials.
“As we move to the new year, we all know that patients are waiting for new medicines today,” says Ben Wolf, chief medical officer at Isomorphic Labs. Wolf joined the company in June and says his focus is advancing Isomorphic Labs into “an industry leading clinical pipeline.” Wolf says their initial focus is oncology and immunology.
Isomorphic Labs Chief Medical Officer Ben Wolf [Photo: Sardari Group for Johns Hopkins University]
“One of the challenges is just the heterogeneic [and] intractable nature of many diseases like cancer. So they’re very challenging to crack,” Wolf says.
What will be imperative for Isomorphic Labs is coupling AI-based approaches to drug development with precision medicine paradigms, i.e. tailoring treatment to specific patients instead of a one-size-fits-all approach.
“Ultimately that will give us the capacity to quickly make safer and more effective medicines for all patients,” Wolf says.
Chris Cuomo is returning to SiriusXM, putting him on the air with morning and evening talk shows.
He will host a two-hour weekday show centered on listener calls on the satellite radio company’s P.O.T.U.S. channel, starting Jan. 20 at 7 a.m. Eastern. Guests on his first few shows include Democratic Maryland Gov. Wes Moore, Republican Kentucky Sen. Rand Paul, sportscaster Bob Costas, and political consultant James Carville, the company said Tuesday.
Cuomo previously hosted a show at the network but quit in 2021, shortly after he was fired from CNN when it was revealed that he assisted his brother, former Democratic New York Gov. Andrew Cuomo, in fighting accusations of sexual harassment.
A former prime-time host at CNN and a reporter at ABC News earlier in his career, Cuomo has reemerged at the cable network NewsNation, where he hosts an evening talk show.
Too much of our politics is about teams instead of truth, said Cuomo, also son of the late Democratic New York Gov. Mario Cuomo. ’Cuomo Mornings’ is going to be about breaking through those barriers, asking better questions and bringing people into the conversation instead of talking past them.
Associated Press
During the pandemic, over two million women left the workforce, many of whom were forced to leave their jobs in the absence of reliable childcare. It took years for female workers to recover from those losses, but eventually the share of working women had surpassed pre-pandemic numbersthough their participation in the labor force still remained lower relative to that of male workers.
In 2025, however, it appears that the gains that women had made in recent years started to slip away.
In the first half of the year, about 212,000 women exited the workforce, and there was a marked dip in employment among working mothers: An analysis by the Washington Post found that the share of working mothers between the ages of 25 and 44 dropped steadily from January to June, resulting in an overall decrease of nearly three percentage points.
The most recent jobs report showed that in December, a total of 81,000 workers left the labor force (which means they are no longer employed or looking for a new job). All of those workers were women, according to a new analysis by the National Womens Law Center (NWLC), which drew on the jobs report data. The overall losses were even higher: 91,000 women left the labor force last month, but that figure was offset by 10,000 men entering the labor force.
Across 2025, there was an increase in the number of women entering the labor forcebut at a much slower clip than in past years. The rate of increase among female workers also pales when compared with that of male workers: The pool of women in the labor force expanded by just 184,000, while the share of men grew by 572,000.
This decline in employment also seems to be having an outsize impact on women of color, according to the NWLC. Unemployment among Black workers has been on the rise in recent months, but particularly for Black women; the unemployment rate for Black women inched up to 7.3% in December, up from 7.1% in November. Latinas also saw a marginal increase in unemployment, from 4.4% the month prior to 4.5% in December. (The overall rate of unemployment, by comparison, has hovered around 4.4%, with an even lower rate for white workers at 3.8%.)
These fluctuations in employment come amid a growing number of hurdles for working mothers. Remote work policies had made it easier for many parents to remain in the workforce, particularly among mothers of young children. Over the last two years, many companies have forced their employees to return to the office, reversing the flexible work arrangements that had enabled parents to juggle their personal and professional obligations. In 2025, leading companies like Amazon and JPMorgan Chase started requiring that workers come into the office five days a week; a report from real estate company Jones Lang LaSalle found that the majority of employees at the hundred largest U.S. companies by revenue were required to come into the office.
The Trump administrations policies and executive actions have also taken a toll on women, and especially working mothers. Trump imposed own return-to-office requirements on federal workers, hundreds of thousands of whom were working remotely. Many of the federal layoffs targeted agencies where women and people of color were overrepresented, and they also disproportionately impacted probationary workersthose in their first year of service or people who have recently been promotedwho are more likely to be women.
Now, working mothers may face new challenges, as Trump takes aim at the childcare programs that make it possible for countless parents to work.
The childcare industry has long struggled with inadequate funding and high labor costs, making it difficult for many providers to keep their doors open. The current administration has exacerbated some of those issues: When he took office, Trump threatened funding for Head Start, which helps subsidize childcare costs for low-income families, and the federal layoffs also compromised the program by cutting staff and making it even harder for underresourced childcare providers to stay afloat. Just days into the new year, Trump has sought to withhold $10 billion in federal funding earmarked for childcare subsidies and social services in Democratic states, due to concerns over alleged fraud. A federal judge has blocked the funding freeze for nowbut theres no telling how Trump may continue to target childcare and caregiving programs, to the detriment of working mothers.
Fewer Americans are signing up for Affordable Care Act health insurance plans this year, new federal data shows, as expiring subsidies and other factors push health expenses too high for many to manage.
Nationally, around 800,000 fewer people have selected plans compared to a similar time last year, marking a 3.5% drop in total enrollment so far. That includes a decrease in both new consumers signing up for ACA plans and existing enrollees re-upping them.
The new data released Monday evening by the Centers for Medicare and Medicaid Services is only a snapshot of a continuously changing pool of enrollees. It includes sign-ups through Jan. 3 in states that use Healthcare.gov for ACA plans and through Dec. 27 for states that have their own ACA marketplaces. In most states, the period for shopping for plans continues through Jan. 15 for plans that start in February.
But even though its early, the data builds on fears that expiring enhanced tax credits could cause a dip in enrollment and force many Americans to make tough decisions to delay buying health insurance, look for alternatives, or forgo it entirely.
Experts warn that the number of people who have signed up for plans may still drop even further, as enrollees get their first bill in January and some choose to cancel.
Health care costs at the center of a fight in Congress
The declining enrollment comes as Congress has been locked in a partisan battle over what to do about the subsidies that expired at the start of the new year. For months, Democrats have fought for a straight extension of the tax credits, while Republicans have insisted larger reforms are a better way to root out fraud and abuse and keep costs down overall. Last week, in a remarkable rebuke of Republican leadership, the House passed legislation to extend the subsidies for three years. The bill now sits in the Senate, where pressure is building for a bipartisan compromise.
Up until this year, President Barack Obama‘s landmark health insurance program had been an increasingly popular option for Americans who don’t get health coverage through their jobs, including small business owners, gig workers, farmers, ranchers, and others.
For the 2021 plan year, about 12 million people selected an Affordable Care Act plan. Enhanced tax credits were introduced the following year, and four years later, enrollment had doubled to over 24 million.
This years sinking sign-ups sitting at about 22.8 million so far mark the first time in the past four years that enrollment has been down from the previous year at this point in the shopping window.
The loss of enhanced subsidies means annual premium costs will more than double for the average ACA enrollee who had them, according to the health care research nonprofit KFF. But extending the subsidies would also be expensive for the country. Ahead of last week’s House vote, the nonpartisan Congressional Budget Office estimated that extending the subsidies for three years would increase the nation’s deficit by about $80.6 billion over the decade.
Americans begin looking for other options
Robert Kaestner, a health economist at the University of Chicago, said some of those who abandon ACA plans may have other options, such as going on a partner’s employer health plan or changing their income to qualify for Medicaid. Others will go without insurance at least temporarily while they look for alternatives.
My prediction is 2 million more people will lack health insurance for a while,” Kaestner said. “That’s a serious issue, but Republicans would argue we’re using government money more efficiently, we’re targeting people who really need it and we’re saving $35 billion a year.
Several Americans interviewed by The Associated Press have said they’re dropping coverage altogether for 2026 and will pay out of pocket for needed appointments. Many said they are crossing their fingers that they aren’t affected by a costly injury or diagnosis.
I’m pretty much going to be going without health insurance unless they do something, said 52-year-old Felicia Persaud, a Florida entrepreneur who dropped coverage when she saw her monthly ACA costs were set to increase by about $200 per month. It’s sort of like playing poker and hoping the chips fall and try the best that you can.”
Ali Swenson and Nicky Forster, Associated Press
As concerns grow over Grok’s ability to generate sexually explicit content without the subject’s consent, a number of countries are blocking access to Elon Musk’s artificial intelligence chatbot. At the center of the controversy is a feature called Grok Imagine, which lets users create AI-generated images and videos. That tool also features a “spicy mode,” which lets users generate adult content.
Both Indonesia and Malaysia ordered that restrictions be put in place over the weekend. Malaysian officials blocked access to Grok on Sunday, citing repeated misuse to generate obscene, sexually explicit, indecent, grossly offensive, and non-consensual manipulated images. Officials also cited “repeated failures by X Corp.” to prevent such content.
Indonesia had blocked the chatbot the previous day for similar reasons. In a statement accompanying Groks suspension, Meutya Hafid, Indonesias Minister of Communication and Digital, said. “The government views the practice of non-consensual sexual deepfakes as a serious violation of human rights, dignity, and the security of citizens in the digital space.”
The responses could be just the beginning of Grok’s problems, though. Several other countries, including the U.K., India, and France, are thinking of following suit.
The U.K. has launched an investigation into the chatbot’s explicit content, which could result in it being blocked in that country as well. “Reports of Grok being used to create and share illegal, non-consensual, intimate images and child sexual abuse material on X have been deeply concerning,” Ofcom, the countrys regulator for the communications services, said in a statement.
Musk, in a social media post following word of the Ofcom investigation, wrote that the U.K. government “just want[s] to suppress free speech.”
Fast Company attempted to contact xAI for comment about the actions in Indonesia and Malaysia as well as similar possible blocks in other countries. An automatic reply from the company read Legacy Media Lies.
Beyond the U.K., officials in the European Union, Brazil, and India have called for probes into Grok’s deepfakes, which could ultimately result in bans as well. (The U.S. government, which has contracts with xAI, has been fairly silent on the matter so far.)
In a press conference last week, European Commission spokesperson Thomas Regnier said the commission was “very seriously looking into this matter,” adding This is not spicy.’ This is illegal. This is appalling. This is disgusting. This is how we see it, and this has no place in Europe.
Musk and X are still feeling the effects of a $130 million fine the EU slapped on the company last month for violating the Digital Services Act, specifically over deceptive paid verification and a lack of transparency in the company’s advertising repository.
Beyond sexualized images of adults, a report from the nonprofit group AI Forensics that analyzed 20,000 Grok-generated images created between Dec. 25 and Jan. 1 found that 2% depicted a person who appeared to be 18 or younger. These included 30 images of young or very young women or girls in bikinis or transparent clothes.
The analysis also found Nazi and ISIS propaganda material generated by Grok.
While the company has not addressed the countries blocking access to its services, it did comment on the use of its tool to create sexual content featuring minors.
“We take action against illegal content on X, including Child Sexual Abuse Material (CSAM), by removing it, permanently suspending accounts, and working with local governments and law enforcement as necessary,” X Safety wrote in a post. “Anyone using or prompting Grok to make illegal content will suffer the same consequences as if they upload illegal content.”
The company has also announced it will limit image generation and editing features to paying subscribers. That, however, likely won’t be enough to satisfy government officials who want to block access to Grok while these images can still be generated.
Advancements in artificial intelligence are shaping nearly every facet of society, including education. Over the past few years, especially with the availability of large language models like ChatGPT, theres been an explosion of AI-powered edtech. Some of these tools are truly helping students, while many are not. For educational leaders seeking to leverage the best of AI while mitigating its harms, its a lot to navigate.
Thats why the organization I lead, the Advanced Education Research and Development Fund, collaborated with the Alliance for Learning Innovation (ALI) and Education First to write Proof Before Hype: Using R&D for Coherent AI in K-12 Education. I sat down with my coauthors, Melissa Moritz, an ALI senior advisor, and Ila Deshmukh Towery, an Education First partner, to discuss how schools can adopt innovative, responsible, and effective AI tools.
Q: Melissa, what concerns you about the current wave of AI edtech tools, and what would you change to ensure these tools benefit students?
Melissa: Too often, AI-powered edtech is developed without grounding in research or educators input. This leads to tools that may seem innovative, but solve the wrong problems, lack evidence of effectiveness, ignore workflow realities, or exacerbate inequities.
What we need is a fundamental shift in education research and development so that educators are included in defining problems and developing classroom solutions from the start. Deep collaboration across educators, researchers, and product developers is critical. Lets create infrastructure and incentives that make it easier for them to work together toward shared goals.
AI tool development must also prioritize learning science and evidence. Practitioners, researchers, and developers must continuously learn and iterate to give students the most effective tools for their needs and contexts.
Q: Ila, what is the AI x Coherence Academy and what did Education First learn about AI adoption from the K-12 leaders who participated in it?
Ila: The AI x Coherence Academy helps cross-functional school district teams do the work that makes AI useful: Define the problem, align with instructional goals, and then choose (or adapt) tools that fit system priorities. It’s a multi-district initiative that helps school systems integrate AI in ways that strengthen, rather than disrupt, core instructional priorities so that adoption isnt a series of disconnected pilots.
We’re learning three things through this work. First, coherence beats novelty. Districts prefer customizable AI solutions that integrate with their existing tech infrastructure rather than one-off products. Second, use cases come before tools. A clear use case that articulates a problem and names and tracks outcomes quickly filters out the noise. Third, trust is a prerequisite. In a world increasingly skeptical of tech in schools, buy-in is more likely when educators, students, and community members help define the problem and shape how the technology helps solve it.
Leaders are telling us they want tools that reinforce the teaching and learning goals already underway, have clear use cases, and offer feedback loops for continuous improvement.
Q: Melissa and Ila, what types of guardrails need to be in place for the responsible and effective integration of AI in classrooms?
Ila: For AI to be a force for good in education, we need several guardrails. Lets start with coherence and equity. For coherence, AI adoption must explicitly align with systemwide teaching and learning goals, data systems, and workflows. To minimize bias and accessibility issues, product developers should publish bias and accessibility checks, and school systems should track relevant data, such as whether tools support (versus disrupt) learning and development, and the tools efficacy and impact on academic achievement. These guardrails need to be co-designed with educators and families, not imposed by technologists or policymakers.
The districts making real progress through our AI x Coherence Academy are not AI-maximalists. They are disciplined about how new tools connect to educational goals in partnership with the people they hope will use them. In a low-trust environment, co-designed guardrails and definitions are the ones that will actually hold.
Melissa: We also need guardrails around safety, privacy, and evidence. School systems should promote safety and protect student data by giving families information about the AI tools being used and giving them clear opt-out paths. As for product developers, building on Ilas points, they need to be transparent about how their products leverage AI. Developers also have a responsibility to provide clear guidance around how their product should and shouldnt be used, as well as to disclose evidence of the tools efficacy. And of course, state and district leaders and regulators should hold edtech providers accountable.
Q: Melissa and Ila, what gives you hope as we enter this rapidly changing AI age?
Melissa: Increasingly, we are starting to have the right conversations about AI and education. More leaders and funders are calling for evidence, and for a paradigm shift in how we think about teaching and learning in the AI age. Through my work at ALI, Im hearing from federal policymakers, as well as state and district leaders, that there is a genuine desire for evidence-based AI tools that meet students and teachers needs. Im hopeful that together, well navigate this new landscape with a focus on AI innovations that are both responsible and effective.
Ila: What gives me hope is that district leaders are getting smarter about AI adoption. They’re recognizing that adding more tools isn’t the answercoherence is. The districts making real progress aren’t the ones with the most AI pilots; they’re the ones who are disciplined about how new tools connect to their existing goals, systems, and relationships. They’re asking: Does this reinforce what we’re already trying to do well, or does it pull us in a new direction? And theyre bringing a range of voices into defining use cases and testing solutions to center, rather than erode, trust. That kind of strategic clarity is what we need right now. When AI adoption is coherent rather than chaotic, it can strengthen teaching and learning rather than fragment it.
Auditi Chakravarty is CEO of the Advanced Education Research and Development Fund.
