The data on workforce development tells a contradictory story. 85% of companies plan to prioritize upskilling their workforce through 2030. At the same time, 63% of employers still identify skills gaps as the single biggest barrier to business transformation.

The explanation for this is that the model most organizations use to develop their people was built for a slower world, and it hasn’t kept up. Learning and development content needs to get scripted, created, reviewed, localized and published.

Even in large, well-resourced organizations, that process can take weeks. By the time most training reaches an employee, the product it was designed to explain has shipped two new updates. The compliance process it covers has been revised. The sales motion it was meant to reinforce has already been changed by the team in the field.

We’ve all had the experience of sitting through mandated corporate training that felt more of a check-box exercise rather than an experience where we actually learn and retain something. To make that learning and development more relevant, companies are changing both the format and the time to delivery.

The Chief Learning Officer’s new mandate

Jayney Howson, Chief Learning Officer at ServiceNow, is working through what a better model actually looks like. ServiceNow University, the company’s initiative to upskill more than three million people by the end of 2027, was recently rebuilt to be AI-native.

The challenge her team faced will be familiar to most L&D leaders: a business shipping AI products on a continuous cycle, a global workforce that needs to stay current and a content production process that couldn’t move fast enough to serve them.

Howson’s response was to rebuild the infrastructure around AI, including AI-generated video, reducing course production time by roughly ten times.

Her team was able to use Synthesia and produce more than 5,000 videos in 18 months, with programs like Sales Academy for their global sales team and partner enablement running consistently and globally. Learning content now reflects what the business is doing today, not what it was doing a few months ago.

According to Jayney, “It feels like a Netflix experience, where it serves up personalized recommendations for each employee. But it can also see that for the job I’m doing right now, the proficiency level I’ve got on a skill is a one and it needs to be a four. So it serves me up that training, too.”

Production is no longer the constraint

ServiceNow’s experience reflects a shift visible across enterprise L&D more broadly. Our research found that 87% of learning professionals are already using AI in their workflows. 72% say the biggest future gain they expect from AI is more personalized learning delivered closer to the moment of need, not just cheaper production.

Those two things have always been linked. Personalization at scale was the stated goal of corporate learning for years, and also its persistent failure. Building individualized learning paths for thousands of employees is not feasible when a single course takes weeks to produce.

When video content can be created, updated and translated in hours, that changes. Programs can be built for specific roles, regions and points in someone’s tenure, rather than averaged out across an entire workforce and useful to no one in particular.

What changes for learning leaders

Organizations that solve the production capacity problem through AI free up their learning function to focus on harder questions.

Which skills actually drive business performance? What does good look like in a specific role, and how do you build toward it? How do you measure whether learning changed behavior, rather than just which employees clicked through a module?

Those are the questions that connect L&D to business outcomes in a way that completion rates never did. The organizations making progress on the skills gap tend to be the ones where learning leaders have been given permission to rethink the operating model, and where AI is being used to close the gap between when knowledge is needed and when it actually arrives.

For Howson, the infrastructure changes matter, but so does the environment around them. She describes her goal for ServiceNow University in terms that go beyond output to making sure the learning experience itself feels like a place where people can take risks.

“We all can remember being a kid and feeling like we were safe,” she said. “This needs to feel like you’re safe to push yourself and not get it right the first time.”

That combination of learning that’s faster, more relevant, and psychologically safe is what separates the organizations closing the skills gap from the ones still trying to solve a 2026 problem with a 2016 model.

Oracle warned customers on Thursday of a critical vulnerability in its PeopleSoft software that hackers have already exploited to breach more than 100 organisations. The flaw, CVE-2026-35273, carries a CVSS score of 9.8 and can be exploited over the internet without any authentication. Oracle has not released a patch.

The advisory came a day after the cybercrime group ShinyHunters claimed responsibility for the mass-hacking campaign. Google’s Mandiant confirmed that the bug Oracle disclosed is the same one ShinyHunters is exploiting. Mandiant said it notified more than 100 global organisations, most of them in the United States.

About two-thirds of the victims are universities and colleges. A ShinyHunters member told TechCrunch the group stole “hundreds of thousands of student records containing full name, home address, phone, email, date of birth, gender, ethnicity, enrollment status, GPA, major, and student ID.” The University of Nottingham was named among the breached institutions.

“While several organizations successfully blocked the activity or remediated the vulnerabilities, others experienced compromise, resulting in stolen data being published on the ShinyHunters Data Leak Website,” Mandiant wrote. Oracle did not respond to TechCrunch’s request for comment.

PeopleSoft is used by large companies and universities to manage payroll, human resources, and student records. The vulnerability affects PeopleTools versions 8.61 and 8.62. ShinyHunters exploited a chain of old and zero-day vulnerabilities to target both cloud and on-premises instances, compromising approximately 300 servers across the 100+ organisations.

The attack follows a pattern. ShinyHunters has spent the past year targeting organisations that share the same vulnerable enterprise software. Previous campaigns hit companies using Salesforce, Gainsight, and education platform Instructure. The group identifies the flaw, finds every company running the software, steals data, and demands a ransom.

Instructure paid the hackers earlier this year after being breached twice. ShinyHunters also defaced the login pages of schools using Instructure’s Canvas portal. The PeopleSoft campaign is the largest yet, and it is ongoing. Oracle recommended mitigations but has not said when a patch will be available.

For any organisation running PeopleSoft, the immediate action is to apply Oracle’s mitigations and restrict internet-facing access to PeopleSoft servers. The broader lesson is one the enterprise software industry keeps relearning: when a critical zero-day hits software used by hundreds of large organisations, the attacker only needs to find it once. AI is making vulnerability discovery cheaper. The defenders patching those flaws are not getting faster. And groups like ShinyHunters are industrialising the exploitation of every window between disclosure and fix.

Oracle warned its corporate customers that there is a critical-rated vulnerability in its PeopleSoft software, which is used by large companies to manage payroll and human resources, a day after a cybercrime group took credit for abusing the flaw as part of a mass-hacking campaign.

The company published the security advisory on Thursday after the hacking group ShinyHunters claimed to have breached more than 100 organizations that use PeopleSoft servers.

Mandiant, the Google-owned security unit that investigates cyberattacks, warned in a blog post that the new Oracle flaw is the same bug that the ShinyHunters group is abusing in its hacking campaign targeting PeopleSoft customers. 

Oracle, which has not released a patch for the vulnerability at the time of writing, said in the advisory that the bug can be exploited over the internet without needing any authentication, such as a password. 

The tech giant recommended that customers who use PeopleSoft software apply its mitigations to prevent exploitation.

On Wednesday, a ShinyHunters member told TechCrunch that the gang compromised the companies by abusing an unpatched flaw in PeopleSoft servers. The bug is known as a zero-day because the company affected, in this case Oracle, had no time to fix it before it was discovered and exploited.

Mandiant confirmed that it has also notified more than “100 global organizations,” most of them in the United States, in an effort to restrict access to their potentially vulnerable systems. The cybersecurity group said that about two-thirds of these organizations are in higher education, which aligns with what ShinyHunters previously claimed.

“While several organizations successfully blocked the activity or remediated the vulnerabilities, others experienced compromise, resulting in stolen data being published on the ShinyHunters [Data Leak Website],” Mandiant wrote. 

Oracle did not respond to TechCrunch’s request for comment. 

The ShinyHunters member told TechCrunch this week that some of the hacked organizations are universities and colleges.

The hacker shared a message they said was sent to one of the victim schools, in which the hackers claimed to have stolen “hundreds of thousands of student records containing full name, home address, phone, email, date of birth, gender, ethnicity, enrollment status, GPA, major, and student ID across all campuses,” among other data. 

PeopleSoft, and its customers, are the latest victims in a long series of hacking campaigns where the ShinyHunters gang targeted organizations that all share the same vulnerable software. 

In the last year, the group targeted several companies that use Salesforce and Gainsight, as well as software provided by education giant Instructure, and among others. 

Once the hackers identify vulnerable software and companies that use it, they try to steal corporate or customer data and then threaten to release it unless the victims pay a ransom. 

Earlier this year, education tech company Instructure said it paid the hackers after they breached the company’s systems twice. As part of the hacking campaign, ShinyHunters defaced the login pages of several schools that use Instructure’s popular school information portal Canvas.

GoPro warned on Monday that there is “substantial doubt about the company’s ability to continue as a going concern.” The action-camera maker reported a 26% revenue decline in Q1 and expects to breach several loan covenants. Shares fell as much as 14%.

The cause is memory. GoPro said its earnings forecast has been “significantly impacted” by an 80% to 115% increase in memory prices. In April, suppliers informed the company of a planned reduction in memory supply that would further reduce forecasted sales. The same DRAM reallocation that is killing the cheap smartphone is now threatening to kill GoPro.

The mechanism is the one we detailed last week. Samsung, SK Hynix, and Micron have redirected wafer capacity from consumer DRAM to high-bandwidth memory for AI data centres. HBM margins run at 70% or higher. Consumer DRAM margins sit between 20% and 30%. The memory makers chose the higher-margin customer. Everyone else pays more or gets less.

GoPro does not have the purchasing power to absorb the price increase. It is not Apple, which can negotiate quarterly contracts and pass costs onto consumers buying $1,000 phones. It is a sub-$1 billion revenue company whose products sell for $300 to $500 and depend on commodity memory to store high-resolution video. When memory costs double, the product becomes unprofitable.

The company has received waivers from its lender after failing to comply with loan covenants. It does not expect to have enough liquidity to meet obligations if default provisions are triggered and outstanding debt becomes due. It has a $50 million second-lien facility from Farallon Capital Management and a revolving credit facility with Wells Fargo as agent.

GoPro has engaged advisors to evaluate strategic alternatives including a potential sale or merger. It is also exploring opportunities in defence and aerospace for “new markets and product categories.” The company already announced plans to cut 23% of its global staff in April.

The defence pivot echoes Faraday Future’s robotics pivot: a consumer electronics company under financial pressure reaching for a higher-margin, government-funded market where the competitive dynamics are different. Whether GoPro’s ruggedised camera expertise translates into defence contracts is unproven.

The only near-term supply relief is coming from China. ChangXin Memory Technologies’ DRAM has been spotted inside Corsair’s retail DDR5 kits. But CXMT is also planning to convert 20% of its capacity to HBM because the margins are irresistible. The consumer memory shortage is structural, not cyclical.

The memory crisis is visible across consumer electronics. The Asus ROG NUC 16 costs $1,200 more than last year’s model, partly due to DDR5 prices. Dell hiked laptop prices 15-20% in December. Apple agreed to pay Samsung a 100% premium on LPDDR5X for the iPhone. These companies can absorb the cost. GoPro cannot.

GoPro was founded in 2002 by Nicholas Woodman. It went public in 2014 at a $3 billion valuation. The company popularised the action camera category and built a brand that became synonymous with extreme sports and adventure content. Its share price peaked above $90 in 2014. It trades below $1 today.

The going-concern warning makes GoPro the most visible corporate casualty of the AI memory reallocation. It will not be the last. Any consumer electronics company with thin margins, limited purchasing power, and dependence on commodity DRAM is facing the same calculus. The AI boom created enormous wealth for three memory makers and the hyperscalers they supply. GoPro is on the other side of that equation.