2026.06.24

Meta Ditches Ray-Ban, AI Embeds Deeper Into the Workplace

Shifts are reshaping hiring, the workplace, and even security audits all at once. After three years of partnership with Ray-Ban, Meta is striking out with its own smart glasses brand. Sakana AI has released Japan’s first commercial multi-agent system. And a new survey shows corporate governance struggling to keep pace with AI adoption, as the “shadow AI” problem quietly intensifies. Today’s dispatch cuts across hiring, operations, security, and hardware to track how AI is rearranging the existing order.

Meta Steps Away from Ray-Ban — Own-Brand Smart Glasses Target Broader Market at $80 Less

For the past three years, Meta and Ray-Ban have been nearly synonymous in the smart glasses market. The EssilorLuxottica partnership was a shrewd move to shed the “gadget-ish” stigma tech companies so often carry and win mainstream design credibility. Now Meta is going it alone — three styles, seven colorways, no Ray-Ban logo.

The price is set $80 below existing Meta Ray-Ban glasses. This isn’t a stripped-down budget option; it’s a clear signal that Meta intends to lower the price floor for smart glasses and reach a wider audience. A Kylie Jenner collaboration colorway is also in the lineup, indicating a deliberate push toward fashion and entertainment consumers.

Meta has also acknowledged privacy improvements are in progress. Meta VP of Wearables Alex Himel stated that “privacy improvements are underway” — a response to the persistent criticism, since the Ray-Ban Meta Glasses launched, that the device can record strangers without their knowledge. The phrase “underway,” however, also reads as deferring disclosure of concrete measures.

The real test of this product is whether Meta can graduate from its dependence on the Ray-Ban brand. Smart glasses still haven’t crossed the threshold of something most consumers actually want to carry around. Lowering the price while deepening Meta AI integration is the right direction, but the fundamental hurdle of camera-equipped wearables and privacy remains. Snapchat charged this market years ago and retreated. With its own-brand bet, Meta is now being asked to prove it can actually win where others couldn’t.

Source: The Verge - Meta launches cheaper smart glasses without Ray-Ban

TikTok-ifying Hiring — Fika Jobs Raises $4M for AI Interview Agents and Short-Form Video Profiles

Stockholm startup Fika Jobs is out to fundamentally rethink the hiring process. The company is building a platform that pairs AI interview agents with short-form video profiles — and it has just raised a $4M pre-seed round.

Here’s how it works: candidates connect their LinkedIn profile, Fika’s AI analyzes their background and generates personalized questions, and candidates then complete a roughly 10-minute video interview with the AI agent, which currently runs on Google’s Gemini models. After the interview, the AI converts responses into short video clips assembled into a profile.

The appeal for recruiters is the “discovery” model. Candidates create a profile once and then wait for companies to find and reach out to them — a shift from the one-way “apply to a job and wait” dynamic to a two-way “post a live profile and let companies come to you” model. The LinkedIn-meets-TikTok framing is accurate, and the appeal to Gen Z candidates is real.

That said, concerns remain. Bias and fairness in AI-driven interviews is a recurring debate in this space, and systems that evaluate video and audio carry particular risk of learning discriminatory patterns. The video format may also amplify existing structural advantages for candidates who are strong verbal communicators. The dependency on Gemini is also worth noting — it accelerates a world where Google holds significant share of AI hiring infrastructure.

Source: TechCrunch - Fika Jobs raises $4M to build a video-first hiring platform where AI agents interview candidates

A survey of 360 workers who use AI on the job returned a striking number: 37.8% said they would continue using AI even if their employer prohibited it. Of those, 19.2% said they would use it personally regardless, and 18.6% said they would push back on the company to keep using it for efficiency reasons. Only 41.1% said they would comply and stop, while 7.2% said they would consider switching jobs or pursuing independent work.

The picture around disclosure is equally stark. Only 18.9% said they report or share all AI tool usage with their employer. The remaining 50.8% fell into categories of “about half,” “some only,” or “almost none.” Companies have no clear view of which AI tools are actually in use inside their own walls — that is the reality for many organizations right now.

Shadow IT has been a long-standing problem, but AI raises the stakes dramatically. Generative AI introduces risks around confidential data inputs, misuse of outputs, and hallucinations. If an employee is drafting a customer-facing email using a personal ChatGPT or Claude account and including client information, the company has no visibility into that risk.

Cybersecurity Cloud, which conducted the survey, points out that “just ban it” no longer works as a management strategy. That assessment is correct. The 7.2% who said they would look for other jobs is also telling — it suggests that the most motivated employees are often the ones most determined to use AI. Companies need to shift from prohibition to visibility and the creation of safe, sanctioned environments for AI use.

Source: ITmedia AI+ - Survey: About 38% of AI users at work say they’d keep using it even if banned

Sakana Fugu Goes Commercial — Japan’s Multi-Agent AI Launches for the Global Market in USD

Sakana AI’s “Sakana Fugu,” released to the public on June 22, is a commercial multi-agent AI system. The design — multiple AI agents coordinated by task, presenting as a single AI model to the end user — is an interesting approach to hiding backend complexity while maintaining scalability.

Pricing comes in three tiers: Standard ($20/month), Pro ($100/month, 10× usage quota), and Max ($200/month, 20× usage quota), with enterprise usage-based billing also in USD. The question “Why USD pricing for a Japanese AI?” has drawn pushback from domestic users. ITmedia asked Sakana AI directly; the answer was that the platform is designed with the global market in mind. A comment that they would “continue to take note of” demand for yen-denominated pricing was as far as the company went.

The technically notable aspect is an architecture that allows the underlying AI models to be swapped out as needed. According to the company, this makes the system less susceptible to disruption from U.S. AI export controls — a consciously geopolitically-aware design choice. Fugu follows the research agent “Marlin” and the “Namazu” model series, and marks Sakana AI’s entry into a genuine monetization phase.

To recap Sakana AI’s background: the company was founded in Tokyo by prominent former Google researchers and raised roughly $200M in its Series A, with NEC and Fujitsu among the investors. The tension between “Japan-born AI startup” as a brand identity and dollar-denominated pricing for a global audience will be an interesting thread to watch as the company matures.

Source: ITmedia - Why is Japan-made AI “Sakana Fugu” priced in USD? Sakana AI says it’s “listening” to demand for yen pricing

NRI Secure Brings Claude Mythos-Level Vulnerability Detection to Market

NRI Secure Technologies launched a commercial vulnerability assessment service on June 23, claiming the ability to detect undisclosed vulnerabilities at a level equivalent to Claude Mythos Preview. Targets include corporate internet-facing servers, core business systems, and software products.

The service combines frontier AI with a proprietary validation framework (harness) developed in-house by NRI Secure. Experts first review source code and software bills of materials (SBOMs) to define the assessment scope. When high-severity vulnerabilities are found, the service provides custom signatures compatible with IPS and WAF systems so that defensive measures can be deployed immediately — designed to close the exposure window between vulnerability discovery and the release and application of a patch.

Claude Mythos has drawn attention as Anthropic’s top-tier model for government and national security use cases, and there are recent reports of SoftBank using it to discover tens of thousands of vulnerabilities. NRI Secure’s claim of “Mythos-equivalent” rests on having reproduced findings that Mythos uncovered using public frontier AI combined with their proprietary harness. This service represents something symbolic: frontier AI capabilities becoming broadly accessible through partners.

Anthropic Japan president Hidetoshi Tojo also commented on the service, describing it as “an effort to detect vulnerabilities ahead of attackers and turn those findings into defenses,” signaling Anthropic’s intent to deepen ties with the NRI Group. AI-assisted security assessment offers dramatic advantages over manual code review in scale, speed, and coverage. Whether this category of service becomes standard practice in corporate security is a trend worth watching.

Source: ITmedia AI+ - NRI Secure launches assessment service detecting undisclosed vulnerabilities “at the level of Mythos”

OpenAI and Anthropic’s FDE Push — Japanese SIers May Not Have Noticed Their New Competition

A report published by Nork Research in May is sounding an important alarm for Japan’s IT industry. Anthropic and OpenAI have each moved to build out a staffing structure designed to embed directly in customer operations — and this is set to have a direct impact on Japanese systems integrator (SIer) businesses.

On May 4, Anthropic established a new company alongside Blackstone and Hellman & Friedman, building out a service capability staffed by “Applied AI Engineers.” On May 11, OpenAI established the “OpenAI Deployment Company (Deploy Co)” alongside TPG, Bain Capital, and SoftBank, prominently featuring a role called “Forward Deployed Engineer (FDE).” The FDE concept was popularized by Palantir and refers to practitioners who embed more deeply in customer operations than traditional consultants or staff augmentation workers. OpenAI has also agreed to acquire UK AI consulting firm Tomoro.

What does this mean for Japanese SIers? Until now, OpenAI and Anthropic were platform providers offering APIs. SIers built solutions on top of those APIs and captured value as the translation layer. If OpenAI and Anthropic are now moving directly into customer operations themselves, the “translator” role that SIers have occupied comes under direct pressure.

Nork Research urges firms to move proactively. Japan’s well-documented preference for the status quo has historically served as a buffer against direct incursion by foreign AI vendors — but that buffer has a time limit. For SIers to remain relevant, they need to internalize not just the ability to “use” AI but the capacity to lead customer business transformation — exactly the role FDEs are being hired to fill. The question facing Japanese SIers is whether to keep competing in a layer that is becoming commoditized, or to move up the stack. That decision is becoming urgent.

Source: ITmedia AI+ - Can the “status quo bias” blocking AI adoption be broken? How OpenAI and Anthropic’s direct operations push will reshape the market

Toyota Finance’s Real-World Playbook — RPA + AI Agents Cut Processing Time from 13 to 4 Minutes

One of the most instructive real-world AI agent success stories right now is Toyota Finance. For customer inquiry email handling, the company cut average processing time per case from 13 minutes to 4 minutes — roughly a 69% reduction.

Their choice was neither “AI agents alone” nor “RPA alone” but a deliberate combination of both. Here’s how it works: RPA bots collect the necessary data from existing systems such as customer information platforms, AI agents then use that data to draft response candidates, and a human performs a final review before sending. The system went live in production in January 2026.

The logic of this design is straightforward. RPA excels at executing repetitive, structured operations reliably — it’s dependable for accessing legacy systems and gathering information. AI agents, meanwhile, can handle non-routine work where content varies case by case, but have limits when it comes to autonomous judgment. Clearly separating the roles lets each compensate for the other’s weaknesses.

At a time when many vendors claim “AI agents will handle everything,” the Toyota Finance case is a useful model of realistic, phased automation. In large Japanese enterprise environments where core systems are numerous and API integration is difficult, the RPA-plus-AI-agent combination is a genuinely practical approach. The fact that processing time dropped to under one-third — even while keeping a human in the loop for final review — demonstrates that “human-machine collaboration” can deliver strong ROI well before full automation is even attempted.

Source: ITmedia AI+ - Why Toyota Finance chose neither “AI agents alone” nor “RPA alone” — and went with both

Takeaway

Across today’s news, one thing is clear: AI has moved past the experimentation phase and is embedding itself into the core of how work gets done. Toyota Finance’s 13-to-4-minute reduction and the 38% shadow AI figure are both evidence of the same thing — AI has taken root on the ground. Meanwhile, the FDE push from OpenAI and Anthropic signals a shift from “selling AI” to “deploying AI” that poses a direct threat to existing SIers and middlemen. On the hardware front, Meta is going independent without Ray-Ban, and Sakana AI has commercialized a domestic multi-agent system aimed at global markets. The center of gravity in the AI industry is moving — steadily, and without pause.