In early 2025, Chinese AI startup DeepSeek gained significant attention after unveiling a large language model that operates with substantially less compute power and cost compared to major U.S. counterparts like OpenAI and Anthropic.
Corgi, the innovative insurance startup, has raised $160 million in its Series B funding round, led by TCV. This investment lifts the company's valuation to $1.3 billion, just four months after completing its Series A.
In April, 28 companies were added to The Crunchbase Unicorn Board, with robotics startups and frontier AI labs dominating the list for the second month running. Two AI labs based in London, both founded by DeepMind researchers, secured massive funding rounds and debuted as unicorns: Ineffable Intelligence with a $5.1 billion valuation and Recursive Superintelligence valued at $4.5 billion. China's ModelBest, specializing in on-device AI models, also joined the unicorn ranks. The robotics sector saw six companies, mostly from China and Japan, gain billion-dollar valuations thanks to innovations like simulated data for robotic intelligence and humanoid robots. Other sectors such as financial services, defense, developer tools, energy, and healthcare added a few unicorns each. Geographically, the U.S. led with 12 new unicorns, followed by China with eight, and smaller numbers from the UK, Germany, Spain, Switzerland, India, and Japan.
NVIDIA has invested upwards of $40 billion in AI equity during the first four months of 2026, with a significant $30 billion stake in OpenAI. The remaining funds have been distributed among CoreWeave, IREN, Corning, Nebius, and about two dozen private investment rounds. This strategy resembles vertical integration more than typical venture capital activity, sparking discussions about potential circular-deal issues.
Uber has long aimed to establish a deeper presence in the autonomous vehicle ecosystem—not just as a ride provider, but as a key player offering data insights, making strategic investments, and functioning as a core platform for distribution. However, its efforts to engage consumers directly remain a critical focus as the industry evolves.
Imagine an observability agent working in a live production environment designed to detect anomalies and take action autonomously. One night, it encounters a spike in anomaly score due to a new, scheduled batch job it has never seen before. Confident in its judgment and with permission to rollback systems, it triggers a rollback that inadvertently causes a four-hour outage. This failure was not because the AI model was flawed but because the system wasn’t tested for unexpected scenarios and boundary conditions beyond its original design.
The major gap today in AI deployment lies in how agents are tested. While identity governance and observability are key focuses, the true challenge is ensuring agents behave correctly when production conditions deviate unexpectedly. Traditional testing methods assume determinism, isolated failure, and clear task completion signals—assumptions that break down with AI agents that behave probabilistically, can propagate complex failures through multi-agent systems, and sometimes signal success even while failing internally. Intent-based chaos testing is proposed to fill this crucial gap by measuring how far agent behavior drifts from its intended purpose under stress, through a structured, multi-phase approach. This method highlights deviations that traditional error metrics miss, allowing enterprises to catch potentially catastrophic failures before deployment and manage risks more effectively.
The approach includes defining behavioral dimensions like tool call patterns, data access boundaries, escalation protocols, and decision timing. An intent deviation score quantifies the degree of departure from expected behavior, guiding actionable responses. Testing progresses gradually from single-component degradations to complex multi-agent failures, calibrated by the risk profile of the deployment. Continuous retesting is necessary as agents evolve, reinforcing this as a governance process—not just a one-time test.
Incorporating intent-based chaos testing as a pre-production gate ensures that autonomous systems are not just functional, but robust against unforeseen real-world challenges, raising the bar for safe AI deployment and reducing costly outages.
Quantinuum announced its plans for a US initial public offering potentially valuing the company at over $20 billion. The company reported nearly $31 million in revenue for the year ending December 31, 2025, alongside a net loss of $192.6 million. Despite significant valuation aspirations, the company’s quantum computer remains in development, leading to cautious investor interest.
Traditional manual penetration testing typically ranges from $10,000 to $50,000, requiring weeks for scheduling and days to perform, with reports that quickly become outdated. London-based cybersecurity firm Intruder, a graduate of GCHQ’s Cyber Accelerator, has introduced AI-powered pentesting agents that mimic human testing methods, conducting comprehensive security assessments in mere minutes.
Emanate, a San Francisco-based AI startup, is transforming the industrial materials sector by developing specialized AI tools tailored for complex sales processes. Unlike broad-spectrum AI companies, Emanate targets manufacturers, distributors, and service providers dealing with materials like steel and metal piping. Their AI quickly generates accurate sales quotes, streamlines customer connections, and identifies new sales opportunities, cutting quote generation time from weeks to almost instantaneous. CEO Kiara Nirghin highlights that recent advancements in AI capability, combined with robust integrations to ERP systems and corporate data, enable these improvements. Emanate’s approach requires a detailed setup process, ensuring systems are customized to client data, which results in high accuracy and measurable revenue growth—aiming for over 40% increase. Initially, AI-generated quotes are human-verified but gradually gain trust, enhancing sales efficiency. The startup’s model supports U.S. manufacturing growth and sustainability through smarter production planning and waste reduction. Emanate envisions expanding its tailored AI technology to other specialized industries like electrical and chemical sectors.
For years, venture capitalists dismissed agencies as mere ‘lifestyle businesses’ unworthy of investment. However, advances in AI have reshaped the industry dynamics, prompting leading VC firms to eagerly back these agencies today.
