Category: 3. Business

Key insights:

• • • Growing pains with expanded AI use — The surge in AI adoption is creating new compliance, governance, and security challenges for businesses, with nearly all firms planning to expand AI use.

    • Many compliance teams are faltering — Current compliance approaches are falling short, with 88% of firms struggling with AI governance and data security, and 37% reporting gaps in their search and e-discovery capabilities.

    • New communications tech tools might help — To address these challenges, organizations need to invest in innovative digital communications governance and archiving platforms that can handle the complexity of AI-generated content and provide forensic-level insight into potential issues.

---

Human-to-AI communication is accelerating across the workplace, creating new compliance, governance, and security challenges for modern organizations. The rise of AI assistants, generative AI (GenAI) tools, and agentic AI introduces a new category of communications — known as aiComms — and a new workplace participant: AI itself. Organizations now face added complexity as they work to manage, monitor, and retain these evolving forms of communication.

Surge in aiComms as firms plan to expand AI use

The results of Theta Lake’s 7th annual Digital Communications Governance Report, based on independent data gathered from 500 IT and compliance leaders in the financial services sector, highlights both the scale of AI expansion and emerging governance challenges.

Nearly all respondents (99%) say their firms plan to implement or expand AI features within their unified communications and collaboration tools, including GenAI assistants (with 92% saying this), AI-powered meeting notetakers and summarization (81%), and customized agentic AI (77%).

Indeed, the widespread adoption of popular GenAI capabilities such as summarization, prompts, and responses is transforming productivity and offering opportunities for firms to realize numerous efficiency gains and improvements. However, the volume of communications is set to accelerate, which will require entirely new approaches to governance, compliance, and content inspection.

In fact, 88% of respondents say their firms already report challenges with AI governance and data security, with the top challenge, identified by nearly half of respondents (47%), is ensuring that AI-generated content is accurate and compliant with regulatory standards. Respondents also cited factors such as difficulties in detecting whether confidential or sensitive data has been exposed in GenAI output (45%), concerns around identifying risky end-user behavior with AI tools (41%), and their ability to track where problematic AI content is shared and with whom (40%).

Fragmented communications compound risk

The report also reinforces the idea that AI is only one part of a broader compliance challenge. Firms are continuing to use multiple unified communications and collaboration platforms, many of which rely on multiple compliance, surveillance, and e-discovery solutions that in turn create inefficiencies, complexity, and gaps in oversight.

Fully 82% of survey respondents say their firms use four or more communication and collaboration platforms. Most organizations also rely on at least three different vendors or repositories for recording, archiving, and supervising communications, the survey shows.

These single-purpose compliance solutions are increasingly inadequate for today’s meshed communications environments, in which text, audio, video, and AI-generated content are created and consumed simultaneously. The limitations of these platforms result in incomplete capture, viewing, and reconciliation, the survey shows, including:

• • • 37% of respondents say their firms report search and e-discovery gaps, up from 31% last year
    • 36% cite surveillance gaps that affect risk detection, remediation, and controls
    • 62% report their firms have difficulties reconstructing and replaying conversations that span multiple communication tools and include textual, voice, visual, and AI-generated interactions

Another continuing concern is the use of off-channel communications, respondents say. Despite more than $4 billion in global fines from government regulators for organizations’ recordkeeping and supervision failures in recent years, more than two-thirds (67%) of respondents say they remain concerned about employees using unmonitored channels.

The findings align with the United Kingdom’s Financial Conduct Authority’s multi-firm review into off-channel communications, released in August, which found that most firms in its sample continue to identify breaches of internal communication policies.

Regulatory expectations

Regulators have made it clear that AI tools are encompassed by existing compliance frameworks. For example, FINRA’s Regulatory Notice 24-09 reminds member firms that:

FINRA’s rules — which are intended to be technology neutral — and the securities laws more generally, continue to apply when member firms use GenAI or similar technologies in the course of their businesses, just as they apply when member firms use any other technology or tool.

This means that highly regulated organizations must be able to confidently enable AI tool while continuing to maintain full oversight of how those tools are used, what they produce, and how their outputs are managed.

Innovative digital communications governance and archiving platforms now include features designed to support AI compliance and security guardrails, helping organizations maintain oversight and transparency in their use of GenAI tools.

Investing in the future

As the report illustrates, confidence in existing compliance approaches remains extremely low at just 2%, down from 8% last year. This underscores the recognition that traditional compliance systems cannot accommodate the volume, diversity, and dynamic nature of today’s digital communications.

As a result, firms are significantly increasing their investment in communications compliance to keep pace with the growing complexity of digital communications, including aiComms. In fact, 86% of respondents say their organizations are already investing more — up from 65% last year — and a further 12% plan to do so.

As human-AI collaboration becomes an integral part of business communication, governance frameworks must evolve to ensure accountability, transparency, and trust. Organizations will need forensic-level insight into AI-generated content to detect potential issues, confirm appropriate use, and respond quickly — all without slowing productivity. Platforms can better support safe GenAI adoption by ensuring AI-generated content aligns with internal policies and regulatory obligations, enabling swift action on risky or non-compliant content, and allowing firms to retain only what’s necessary to meet oversight and recordkeeping requirements.

---

In the future, tiny flying robots could be deployed to aid in the search for survivors trapped beneath the rubble after a devastating earthquake. Like real insects, these robots could flit through tight spaces larger robots can’t reach, while simultaneously dodging stationary obstacles and pieces of falling rubble.

So far, aerial microrobots have only been able to fly slowly along smooth trajectories, far from the swift, agile flight of real insects — until now.

MIT researchers have demonstrated aerial microrobots that can fly with speed and agility that is comparable to their biological counterparts. A collaborative team designed a new AI-based controller for the robotic bug that enabled it to follow gymnastic flight paths, such as executing continuous body flips.

With a two-part control scheme that combines high performance with computational efficiency, the robot’s speed and acceleration increased by about 450 percent and 250 percent, respectively, compared to the researchers’ best previous demonstrations.

The speedy robot was agile enough to complete 10 consecutive somersaults in 11 seconds, even when wind disturbances threatened to push it off course.

“We want to be able to use these robots in scenarios that more traditional quad copter robots would have trouble flying into, but that insects could navigate. Now, with our bioinspired control framework, the flight performance of our robot is comparable to insects in terms of speed, acceleration, and the pitching angle. This is quite an exciting step toward that future goal,” says Kevin Chen, an associate professor in the Department of Electrical Engineering and Computer Science (EECS), head of the Soft and Micro Robotics Laboratory within the Research Laboratory of Electronics (RLE), and co-senior author of a paper on the robot.

Chen is joined on the paper by co-lead authors Yi-Hsuan Hsiao, an EECS MIT graduate student; Andrea Tagliabue PhD ’24; and Owen Matteson, a graduate student in the Department of Aeronautics and Astronautics (AeroAstro); as well as EECS graduate student Suhan Kim; Tong Zhao MEng ’23; and co-senior author Jonathan P. How, the Ford Professor of Engineering in the Department of Aeronautics and Astronautics and a principal investigator in the Laboratory for Information and Decision Systems (LIDS). The research appears today in Science Advances.

An AI controller

Chen’s group has been building robotic insects for more than five years.

They recently developed a more durable version of their tiny robot, a microcassette-sized device that weighs less than a paperclip. The new version utilizes larger, flapping wings that enable more agile movements. They are powered by a set of squishy artificial muscles that flap the wings at an extremely fast rate.

But the controller — the “brain” of the robot that determines its position and tells it where to fly — was hand-tuned by a human, limiting the robot’s performance.

For the robot to fly quickly and aggressively like a real insect, it needed a more robust controller that could account for uncertainty and perform complex optimizations quickly.

Such a controller would be too computationally intensive to be deployed in real time, especially with the complicated aerodynamics of the lightweight robot.

To overcome this challenge, Chen’s group joined forces with How’s team and, together, they crafted a two-step, AI-driven control scheme that provides the robustness necessary for complex, rapid maneuvers, and the computational efficiency needed for real-time deployment.

“The hardware advances pushed the controller so there was more we could do on the software side, but at the same time, as the controller developed, there was more they could do with the hardware. As Kevin’s team demonstrates new capabilities, we demonstrate that we can utilize them,” How says.

For the first step, the team built what is known as a model-predictive controller. This type of powerful controller uses a dynamic, mathematical model to predict the behavior of the robot and plan the optimal series of actions to safely follow a trajectory.

While computationally intensive, it can plan challenging maneuvers like aerial somersaults, rapid turns, and aggressive body tilting. This high-performance planner is also designed to consider constraints on the force and torque the robot could apply, which is essential for avoiding collisions.

For instance, to perform multiple flips in a row, the robot would need to decelerate in such a way that its initial conditions are exactly right for doing the flip again.

“If small errors creep in, and you try to repeat that flip 10 times with those small errors, the robot will just crash. We need to have robust flight control,” How says.

They use this expert planner to train a “policy” based on a deep-learning model, to control the robot in real time, through a process called imitation learning. A policy is the robot’s decision-making engine, which tells the robot where and how to fly.

Essentially, the imitation-learning process compresses the powerful controller into a computationally efficient AI model that can run very fast.

The key was having a smart way to create just enough training data, which would teach the policy everything it needs to know for aggressive maneuvers.

“The robust training method is the secret sauce of this technique,” How explains.

The AI-driven policy takes robot positions as inputs and outputs control commands in real time, such as thrust force and torques.

Insect-like performance

In their experiments, this two-step approach enabled the insect-scale robot to fly 447 percent faster while exhibiting a 255 percent increase in acceleration. The robot was able to complete 10 somersaults in 11 seconds, and the tiny robot never strayed more than 4 or 5 centimeters off its planned trajectory.

“This work demonstrates that soft and microrobots, traditionally limited in speed, can now leverage advanced control algorithms to achieve agility approaching that of natural insects and larger robots, opening up new opportunities for multimodal locomotion,” says Hsiao.

The researchers were also able to demonstrate saccade movement, which occurs when insects pitch very aggressively, fly rapidly to a certain position, and then pitch the other way to stop. This rapid acceleration and deceleration help insects localize themselves and see clearly.

“This bio-mimicking flight behavior could help us in the future when we start putting cameras and sensors on board the robot,” Chen says.

Adding sensors and cameras so the microrobots can fly outdoors, without being attached to a complex motion capture system, will be a major area of future work.

The researchers also want to study how onboard sensors could help the robots avoid colliding with one another or coordinate navigation.

“For the micro-robotics community, I hope this paper signals a paradigm shift by showing that we can develop a new control architecture that is high-performing and efficient at the same time,” says Chen.

“This work is especially impressive because these robots still perform precise flips and fast turns despite the large uncertainties that come from relatively large fabrication tolerances in small-scale manufacturing, wind gusts of more than 1 meter per second, and even its power tether wrapping around the robot as it performs repeated flips,” says Sarah Bergbreiter, a professor of mechanical engineering at Carnegie Mellon University, who was not involved with this work.

“Although the controller currently runs on an external computer rather than onboard the robot, the authors demonstrate that similar, but less precise, control policies may be feasible even with the more limited computation available on an insect-scale robot. This is exciting because it points toward future insect-scale robots with agility approaching that of their biological counterparts,” she adds.

This research is funded, in part, by the National Science Foundation (NSF), the Office of Naval Research, Air Force Office of Scientific Research, MathWorks, and the Zakhartchenko Fellowship.

Stay informed with free updates

Simply sign up to the Technology sector myFT Digest — delivered directly to your inbox.

The writer is co-founder of BlankPage Capital, co-founder of Graphcore and author of ‘How AI Thinks’

It was a small company based in London that started the race to build the world’s most advanced artificial intelligence. DeepMind, founded in 2010, pioneered deep-learning AI and is working towards human-level AI. But its acquisition by Google in 2014 means that the company’s breakthroughs are largely benefiting the US, not the UK. As the AI race intensifies, the UK needs to find a way to build its own large-scale tech companies.

The usual conversation is about ways that the country can bridge the funding gap and encourage innovation. But this is not where the problem lies. The UK is world class at creating technology companies, often via spinouts from our universities. In 2023, according to figures provided by database company Dealroom to investor Phoenix Court, Bay Area seed and early-stage companies (those raising between $1mn and $15mn in funding) raised $4.5bn. The equivalent companies in the UK raised $4.1bn. So the UK is not far behind at the earliest start-up stage. This first link in our innovation ecosystem is not broken.

Yet by the final start-up stage, where these companies need to raise $100mn or more in funding to drive growth, US businesses are way ahead, raising five times as much as their UK counterparts. What is it that goes wrong in between?

The answer lies in the all-important middle stage, known as early growth, when companies have built a team and a product and are trying to expand into global businesses. In 2023, Bay Area early-growth stage companies received nearly two times the funding of their UK counterparts — at just under $14bn compared to just over $7bn.  

DeepMind was passing through exactly this growth stage when the co-founders decided to sell their business and gain access to the resources that were available as a division of Google. Last year, I sold my AI chipmaker Graphcore to Japan’s SoftBank for exactly the same reason. 

UK policymakers are focused on funding, pouring energy into reforming capital markets, corralling pension funds to invest in UK companies and putting more state funds into venture capital — including the new £500mn Sovereign AI Unit announced in July.  

This is all helpful. But the answer isn’t money — or rather, it isn’t just money. Yes, access to capital is critical. But to attract more conservative, late-stage, global private capital, UK tech companies must be mature enough to show commercial traction, the potential for massive revenue growth and the ability to generate future profits. What is holding them back is a lack of the right kind of support to connect their businesses to customers and to turn them from technology start-ups into true commercial entities.

If your company has access to Silicon Valley investors and your early-growth stage funding is led by Benchmark Capital, Andreessen Horowitz or one of the other leading tech VC firms, you get more than money. These firms will provide access to senior leaders at the biggest tech companies, many of which they funded. You will receive help from partners who built their own global businesses and are highly connected to the market.

This is the type of support we must offer in the UK by replicating the Silicon Valley ecosystem of specialised support for early-growth companies. Founders who have built their own tech companies must pass on their experience. VCs must double down on this part of the innovation ecosystem and help early-growth tech businesses win deals in international markets and connect with expert mentors who have scaled and exited major companies.  

The UK has the potential to lead in AI, just as we do in fintech. But to do so, we need to build a new base of VCs. It is their expertise that will help our most promising tech companies find global success.

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

They grow up so fast. Anthropic, a maker of artificial intelligence models and rival to OpenAI, has hired lawyers ahead of an initial public offering that could value it at $350bn next year. By that point the company would have reached the not-at-all-grand age of five.

That makes it a pretty good example of AI companies’ turbocharged growth. As a comparison, Google went public six years after its founding, achieving a valuation of about $23bn. Facebook took eight years to spring into public markets with roughly a $100bn price tag. The geriatric Microsoft waited 11 years and debuted in 1986 at around $800mn.

Behind the hype, at least, is a business. Anthropic’s main product is its chatbot Claude. That generates revenue, if not yet profit: Anthropic has projected that it could make $70bn in sales by 2028, The Information has reported. That would put its mooted valuation at five times that sum. Meta went public in 2012 at, with hindsight, a multiple of six times its three-year-hence sales; China’s Alibaba at seven times and Palantir at 10.

When investors do get the chance to buy stock in Anthropic directly, joining existing backers Amazon, Google, Microsoft and Nvidia, they will be benchmarking it particularly closely with OpenAI, the maker of ChatGPT, whose latest valuation of $500bn is also five times 2028 projections.

Who wins the bake-off depends on what flavours an investor prefers. Anthropic seems more popular with companies, with a 32 per cent share of the “enterprise” market as of the end of July, according to Menlo Ventures — which it should be noted is an Anthropic investor. That’s helpful, because businesses are more likely to pay up for AI than consumers.

Anthropic is also a more narrow business. It builds models, and that’s about it. OpenAI, meanwhile, is investing in data centres, pocket-sized devices, other companies’ shares and its own web browser. Some might call that sprawl; a venture capitalist might be more likely to call it “full stack”. Seen that way, Anthropic might more resemble a Palantir or a Salesforce, where OpenAI has shades of Google parent Alphabet or Microsoft.

Perhaps the toughest thing to value is what Anthropic might see as its greatest asset: its principles. The company was founded to be a safer alternative to OpenAI, building bots that are “helpful, honest and harmless”. Indeed, the Center for AI Safety deems Anthropic’s products the least likely among major models to “overtly lie” or furnish answers to “hazardous expert-level virology queries”.

Whether investors will pay a premium for that — or instead demand a discount — remains to be seen. In the meantime, much can happen. By the time it goes public, if it does, AI may have taken another leap forward, or tripped on its own hype. It’s therefore worth thinking about how Anthropic might justify a $350bn valuation — while also being prepared to tear those assumptions up and start again.

john.foley@ft.com

Unlock the White House Watch newsletter for free

Your guide to what Trump’s second term means for Washington, business and the world

Bond investors have told the US Treasury they are concerned about Kevin Hassett’s potential appointment as Federal Reserve chair, worrying he will cut interest rates aggressively to please President Donald Trump.

The Treasury department solicited feedback on Hassett and other candidates in one-on-one conversations with executives at major Wall Street banks, asset management giants and other big players in the US debt market, according to several people familiar with the conversations.

The discussions took place in November, before Treasury secretary Scott Bessent held his second round of interviews with candidates to replace Jay Powell as Fed chair when his term expires in May 2026, these people said.

A White House spokesperson said “the president will continue to nominate the most qualified individuals to the federal government, and until an announcement is made by him, any discussion about potential nominations is pointless speculation”.

The Treasury declined to comment.

Hassett, the White House’s top economic official, has emerged as a frontrunner for the position in recent weeks, as Trump and Bessent have whittled down the list of potential candidates from 11 initial contenders.

Trump on Tuesday said he planned to name his pick for Fed chair “early” next year, and signalled Hassett was a “potential” contender. The US dollar briefly slipped on the president’s mention of Hassett.

The doubts about Hassett reflect a broader anxiety on Wall Street about the transition at the Fed’s helm as Trump prepares to nominate a new leader of the central bank. Some senior bond market participants would have preferred other candidates such as BlackRock’s Rick Rieder and Fed governor Christopher Waller who were seen as more independent from Trump than Hassett.

Several of the market participants the Treasury contacted said they were worried about Hassett’s alignment with Trump, who has insisted rates should be cut sharply and has called Powell a “stubborn mule” for the central bank’s decision to only modestly lower borrowing costs this year.

The bankers and investors were worried that Hassett could agitate for indiscriminate rate cuts even if inflation continues to run above the Fed’s 2 per cent target, according to three people familiar with the conversations.

“No one wants to get Truss-ed,” said one market participant, referring to the shock in the UK bond market in 2022 triggered by the then-prime minister Liz Truss’s plans for unfunded tax cuts.

The prospect of a dovish Fed chair was viewed as particularly worrisome to the big bond managers in the event that inflation in the US rises next year. The Fed’s preferred inflation gauge registered 2.7 per cent in August.

The combination of loose monetary policy and higher inflation could ignite a sell-off in long-term Treasuries, said one market participant.

Some market participants were also unconvinced Hassett would be able to win round a divided Fed board and corral consensus on rate decisions, the people added. 

Among the array of participants in those conversations were members of the group of Wall Street bond titans who make up the Treasury Borrowing Advisory Committee, which counsels Bessent on market and issuance questions, according to two people familiar with the matter.

When Hassett, a career economist whose work has focused on tax policy, met with TBAC earlier this year, he spent little time talking about markets, instead pitching White House priorities, including a discussion about Mexican drug cartels, these people said.

A Washington insider, Hassett served as a senior economic adviser on the presidential campaigns of John McCain, George W Bush and Mitt Romney, before joining the White House during Trump’s first term as chair of the Council of Economic Advisers.

He also worked at the conservative think-tank the American Enterprise Institute — and at the Fed, where staffers who worked with him remember him as ambitious. 

Robert Tetlow, a senior policy adviser who recently left the Fed, said Hassett struck him as “smart, eloquent and self-assured”. 

However, concerns abound that his closeness to a president who has spent the past year attacking the US central bank will threaten the institution’s independence. 

“Kevin Hassett is more than capable of doing the job of Fed chair, it’s just a question of who shows up,” said Claudia Sahm, a former Fed economist who is now chief economist at New Century Advisors. “Is it the Kevin Hassett who is the active participant in the Trump administration? Or Kevin Hassett the independent economist?” 

John Stopford, head of managed income at asset manager Ninety One, added: “I think the market sees him as a Trump stooge which erodes the Fed’s credibility at the margin.”

Additional reporting by Ian Smith in London

Liverpool City Council

Liverpool City Council