BETA
This is a BETA experience. You may opt-out by clicking here
Edit Story

Under Pressure: 25% Of Tech Leaders Say They Moved Too Fast On AI

Plus: This Deep Sea Mining Company Wants To Scoop Metals For EV Batteries Off The Ocean Floor

This is the published version of Forbes’ CIO newsletter, which offers the latest news for chief innovation officers and other technology-focused leaders. Click here to get it delivered to your inbox every Thursday.


Two miles below the surface of the Pacific Ocean, in the expansive sea south of Hawaii, sits a vast quantity of nickel, manganese, cobalt and copper—all vital to the manufacture of lithium-ion batteries.

Whether we’re talking electric vehicles, or consumer electronics more broadly, the supply chain for the batteries that will power lower-carbon energy in the future depends on these metals.

If you ask Gerard Barron, CEO of deep sea mining operation The Metals Company, billions of palm-sized metal nodules are in fact just sitting on the seabed at the bottom of the ocean, waiting to be scooped up. But of course, it’s not that easy.

Barron has solved for a few of the obstacles in his way: Certainly no human diver is going that deep in the ocean, so a robotic nodule collector machine combs the ocean floor, vacuuming the metal bounty up to a control ship at the surface. While seafloor harvesting does have environmental critics, a half-dozen studies have shown that impact to sea life is minimal given how little life the “abyssal plain” can sustain in the first place.

That leaves two significant blockers: bureaucracy, and the business itself. Legally, who is allowed to claim and pick up underwater rocks? There is a 1994 United Nations charter that governs marine activity in international waters, but the U.S. Senate has not ratified it. As other nations move in to capitalize on deep sea minerals, namely China, there’s a push to make U.S. involvement more formal.

Then there’s the pesky problem of financing a company through all the aforementioned obstacles: The Metals Company has no revenues, and its share price has fallen 80% since it went public in 2021 with a $2.9 billion valuation. As of the last quarter, the company was down to its last $25 million. A couple months ago the company was hoping to commence commercial nodule collection in 2025, but given rulemaking bureaucracy, they now say it will be more like mid-2026 before they can deploy.

Like so much of the progress modern technology has enabled, there are going to be real-world barriers. It’s at least part of why the 2010s era adage of “move fast and break things” has, more or less, been reduced to a meme at this point.

CYBERSECURITY

After spending 15 years in Israeli intelligence and seven in the private sector, Sanaz Yashar has raised $30 million for a new startup called Zafran, which aims to prevent spies and cybercriminals from exploiting known vulnerabilities to break into companies’ networks, using technologies the businesses already have. Of course, she’s targeting a pressing problem: The average data breach costs the victim company $4.5 million, according to IBM data from 2023, and previous studies have shown cyberattacks costing the global economy hundreds of billions every year.

After determining which vulnerabilities are most pressing, Zafran scans a customer’s network and probes APIs to look for which controls can fix a given weakness, then translates that into something even a non-technical executive can understand, says Yashar. ​​“It’s almost biology, it’s like a self-healing platform,” she says, explaining that the product looks at the body of each customer to determine how it can best repel infection.

On the government side, the U.K. is increasingly confident it was hacked by a Chinese state-affiliated entity between 2021 and 2022, and has formally accused the country of cyber campaigns against members of Parliament and the U.K. Electoral Commission. The personal data of millions of voters was stolen, and email accounts of MPs who were members of the Inter-Parliamentary Alliance on China were also targeted, though U.K. officials say they caught and mitigated the attack before any accounts were compromised. At the time of the breach, the attack was widely attributed to Russia.

ARTIFICIAL INTELLIGENCE

A new open source LLM from Databricks aims to help other companies build and train their own AI agents. Called DBRX, the model will allow businesses to create custom artificial intelligence based on their own proprietary data. Companies will be able to use it for free, but must pay if they want to use it with the Databricks cloud platform. The company charges pay-as-you-go fees for its tools, with discounts for committed usage.

“We think that open source just moves the industry forward. Companies don’t want to get locked into different proprietary standards,” CEO Ali Ghodsi said during a briefing with reporters on Monday. “So it’s actually also a competitive advantage for us business-wise.”

On the exploratory (and potentially unsettling) front of AI, New York-based startup Hume claims its model can detect more than 24 distinct emotional expressions lacing a person’s voice—from nostalgia to awkwardness to anxiety—and respond to them accordingly. The “emotionally intelligent” conversational AI uses an in-house model to interpret emotional tone, and for more complex content, it relies on external LLMs, including OpenAI’s GPT 3.5, Anthropic’s Claude 3 Haiku, and Microsoft’s Bing Web Search API generates responses within 700 milliseconds.

Forbes’ Rashi Shrivastava writes that Hume’s tech also occasionally pauses when speaking, and will even chuckle—slightly disconcerting to hear coming from a computer. We’re definitely getting “I’m sorry Dave, I’m afraid I can’t do that” vibes…

FACTS + COMMENTS

When it comes to AI integration, some IT leaders say they regret jumping in too soon, according to a survey from Asana’s Work Innovation Lab.

1 in 4: Portion of tech leaders who already regret investing in AI too quickly

72%: Share of leaders who reported experiencing stress at least monthly over pressure to implement AI-driven automation, over the last six months

14%: Share of leaders who say their organization has a definitively allocated budget for AI initiatives

STRATEGIES + ADVICE

Researchers believe they’ve found the first example of cyberattacks exploiting AI computing vulnerabilities, in this case, through the popular open source software called Ray. Here’s more on the breach and broader risk.

How can you get the most creativity and innovation from the employees you already have? Empower a team of ‘intrapreneurs.’

QUIZ

A signed business card from the mid-1980s recently sold at auction for a whopping $181,183. Whose card was it?

A. Steve Wozniak

B. Steve Jobs

C. Bill Gates

D. Linus Torvalds

See if you got the answer right here.