GOVERNANCE

 AI warning from intel agencies should be heeded

AI may soon be capable of completely dismantling our cybersecurity defenses. 

In a joint statement, the Five Eyes cybersecurity alliance, which includes intelligence agencies from the US, UK, New Zealand, Australia and Canada, called for leaders to "act now" on AI's cyber threats. The statement warned that AI models capable of taking down businesses and governments are mere months away. 

The agencies said that frontier models are anticipated to "exceed current industry expectations," fundamentally transforming both cyber offense and defense. While the tech provides cyber defenders with a new arsenal of tools to better protect systems, it also enables malicious actors to make their attacks faster and more complex and shrinks the window between discovering a vulnerability and acting on it. "The timeline is not years, it is months," the agencies said. 

Five Eyes urged business and government leaders to preempt these threats in several ways: 

• First, understand and assess the current risk, as well as the current state of preparedness, and be flexible to adapt to the quickly evolving threat landscape

• Prioritize foundational cybersecurity best practices, including reducing attack surfaces, implementing strong identity controls, addressing vulnerabilities in legacy systems and accelerating the patching process

• Give cyber leaders the authority and resources to defend properly

"It is not enough to have controls," the statement said. "Leaders must be confident those controls will perform during a real incident. This requires reassessing long-standing trade-offs and using AI deliberately to strengthen defense — not just improve efficiency." 

Now, the race is on for organizations to solidify their security fundamentals and prepare for "machine-speed" attacks, Erik Avakian, technical counselor at Info-Tech Research Group, told The Deep View. This means that areas such as backups, asset inventory, zero trust, vulnerability management, and incident response need to be at a mature level. 

"Now is the time to start exploring automation where it makes sense," said Avakian. "AI speed and agility are too fast now and will only get faster. It’s going to end up forcing defenders to embrace more automation."

The warning comes amid heightened fears about AI's cybersecurity capabilities, especially following the release (and subsequent US government muzzling) of Anthropic's Fable 5 and Mythos 5, the most powerful models in its lineup thus far. Anthropic itself said the Mythos Preview could quickly find software exploits, creating opportunities for attackers and other bad actors. 

Rival OpenAI, meanwhile, is putting out its own offerings targeting cyberdefenders, including a slew of new tools as part of Project Daybreak debuted on Monday. OpenAI's line-up includes an update to the Codex Security plugin, an expanded release of GPT‑5.5‑Cyber, a partner program to give organizations more access, and an initiative called "patch the planet," which helps find and fix vulnerabilities in popular open source projects.   

But solving this problem may extend beyond just any one model provider, organization or enterprise, said Avakian. 

"I believe strongly that this is a shared responsibility," he said. "Governments and AI providers both certainly have responsibilities around guardrails, standards, and safety. But business organizations can't just outsource accountability for their own cyber risk."

One of the biggest cybersecurity vulnerabilities an organization can suffer from is hubris. A business, enterprise or agency may think that their legacy systems will hold up against new forms of attack, that they can't afford to invest in cybersecurity for fear of falling behind on innovation or that they simply aren't a likely target for attackers. Those organizations, however, are at higher risk of being walloped. As Five Eyes warns, exceedingly powerful AI raises the bar for everyone: Amateur hackers now have the capability to complexify their simple attacks, while sophisticated malicious actors now have the ability to scale their campaigns to immeasurable heights. With that kind of ammo, everyone is a target. 

TOGETHER WITH MERCURY

Eliminate the “switching tax” on your finances

Every finance task takes multiple steps across multiple tools. Mercury Command inverts that.

It's AI built directly into your Mercury account that surfaces insights and completes financial work based on your live data, with you approving every action before anything moves.

Say "send that contractor payment" or "freeze that card" and it's done.

No switching tools, no navigating dashboards, no stitching together context from three different places. Just direct Command.

Try Mercury Command →

WORKFORCE

Why workers are burning out from AI pressure

AI promised to make work easier. Instead, some leaders fear it's making workers uneasy.

During a panel on AI and wellness at VivaTech in Paris, Europe’s largest tech conference, business leaders and a psychiatrist argued that AI is increasing workplace stress by adding new demands, fueling uncertainty, and challenging workers' sense of purpose.

The pressure to adopt AI has increased workloads and complicated decisions about when and how to use the technology, the panelists said, leaving workers feeling confused, anxious, and exhausted. Nine out of 10 U.S. workers are experiencing cognitive and mental strain, according to a recent study from mental health app Headspace. More than a third say it has worsened over the past year, in part because of AI. 

"Everybody's being told, 'I don't know what's going to happen with your job,'" said Tom Pickett, CEO of Headspace. "You have to relearn everything, and what you learn is no longer relevant. The result is that people are becoming increasingly burned out at work."

That burnout is exacerbated by the pace of technological change, said Dr. Fanny Jacq, psychiatrist and chief medical officer at Eutelmed, a French mental healthcare provider. Rapid advancements in AI are outpacing people's ability to adapt, creating a feeling of uncertainty among workers as they navigate their careers. 

"Human beings like to feel in control," Dr. Jacq said. "When people don't know what will happen to their job, their skills, and their professional identities, anxiety naturally increases."

Panelists also challenged the assumption that AI automatically reduces stress by eliminating work. Pickett said creating more documents and reviewing AI-generated outputs can introduce new distractions. Dr. Jacq added that time saved by AI can be replaced with more meetings and tasks, potentially increasing workloads while eroding critical thinking skills.

As AI transforms jobs, employees may begin to question their sense of purpose, according to Emily Witko, head of culture at AI platform Hugging Face. Many workers derive a sense of identity from their jobs, and automating core responsibilities can leave them, as Witko put it, "floating around in space" and wondering where they fit.

"If you're missing meaningful work, I think that can actually really degrade your well-being," Witko said.

To mitigate AI's impact on mental health, panelists said companies need to deploy the technology more intentionally. That means giving workers clear guidance on when and how to use AI, redesigning workflows around the technology, and preserving human skills such as critical thinking, judgment, empathy, and trust.

The utopian vision of AI supercharging productivity is being tempered by reality. Companies are racing to deploy AI in pursuit of efficiency and cost savings. But that doesn't automatically bode well for workers and their well-being. As employees scramble to adapt to new tools and expectations, AI can create new problems instead of eliminating old ones. That raises a fundamental question: Can AI improve work if the underlying processes are already broken? As Hugging Face's Witko put it, "If you have a crappy, frustrating process to begin with, just adding an AI tool isn't going to solve the problem."

TOGETHER WITH CHECKSUM

Don't stop at CI/CD. Ship faster with continuous verification.

63% of engineering teams now ship code faster with AI. 72% have already had a production incident from AI-generated code. The bottleneck didn't disappear. It moved downstream.

Checksum is an AI-native continuous testing platform that auto-generates and self-heals your E2E test suite, runs inside your existing CI/CD pipeline, and keeps pace with the velocity you're already getting from AI coding agents.

Clearpoint Strategy saves $500k a year. Postilize ships 30% faster.

Setup takes under 30 minutes.

STARTUPS

One startup's quest to make a DNA language model

We typically think of AI models as built around language. But one startup is creating models built around the "grammar of our DNA." 

Radical Numerics, an AI research lab spun out from Stanford University on the quest to build "general biological intelligence," emerged from stealth last week with $50 million in seed funding. Additionally, the company announced its latest model: Omnii, a "genome language model" built for researchers and biologists, in research preview. 

"Most of the DNA in the human genome, we still don't actually understand what it does. We don't understand the grammar, so to speak, of our DNA," Eric Nguyen, CEO and co-founder of Radical Numerics, told The Deep View. "We thought this was a way to accelerate our understanding of disease, and hopefully to make treatments to cure disease." 

Though the company has only just revealed itself to the public, it's been at work for almost a year, said Nguyen, and has already created and released models. 

• Evo and Evo 2, its first two models, are the first AI models capable of both reading and writing DNA at scale. The project is entirely open source, Nguyen said, and was used to design novel CRISPR systems and create the first complete AI-designed genome, called a bacteriophage. 

• Because the field is so nascent, the company released these models open source as a means of pushing innovation and getting it in the hands of researchers "in an open way," he said. "We felt that we had to show people the recipe and make it open source to actually drive adoption." 

However, amid a growing discussion of AI's potential use in developing bioweapons, Nguyen said that the company approaches its technology in a "dual prong" way, considering the tech for both biological design and biological defense. It's why the company decided to keep Omnii proprietary, compared to its Evo predecessors: Being able to control who has access to its models is vital.  "We don't want this to be an academic toy anymore," Nguyen said. 

"We realized that this technology has this tremendous potential, obviously, to understand human health, but it can be potentially misused to create something very dangerous," he added.

Radical Numerics joins a growing movement targeting bioscience as a use case for AI. OpenAI introduced GPT-Rosalind specifically for life sciences research and drug discovery, and Anthropic has its own version of Claude for Life Sciences. Additionally, each of these initiatives has a precedent: Alphafold, Google DeepMind's AI project predicting the 3D structures of proteins, which won the Nobel Prize in 2024. And there's a reason this particular niche is so attractive — One of the big pieces of the utopian vision for AI is to provide researchers with the tools to exponentially accelerate scientific discovery, positioning the tech as a panacea for all of the hurdles that ail researchers. Still, Radical Numerics acknowledges the biggest risk: That a tool of such power can be used to tip the scales in a dangerous direction, if not controlled. 

LINKS

• OpenAI, Getty Images announce licensing deal for images in ChatGPT

• Micron, Anthropic ink strategic partnership to scale AI infrastructure

• JD CEO says delivery workers will "sooner or later" be replaced by robots

• Google invests $75 million in studio A24, forges AI film tools partnership

• SpaceX, Reflection sign computing deal worth up to $6.3 billion

• Chip firm Groq raises $650 million to expand data center capacity

• Sakana Fugu: An LLM trained to call on other models in an agent cool, designed to orchestrate other models in complex, multi-step tasks. 

• Grok Imagine: xAI's flagship model now has multitasking agents.

• HappyHorse 1.1: Alibaba's latest model for production-ready video synthesis.

• ElevenLabs' Ad Engine: Users can now connect Google, Meta, and LinkedIn ad accounts to ElevenLabs

• Anthropic: Applied AI Claude Evangelist, Startups 

• Google: Image Processing Engineer

• Motorola Solutions: Senior Machine Learning Engineer

• Cognition: Research, Post-Training

GAMES

POLL RESULTS

Would you purchase a pair of AR glasses?

Yes (20%)
Maybe, but not yet (35%)
No (39%)
Other (6%)