Top Tech News Today, August 18, 2026: Apple, Baidu, ByteDance, Google, Meta, OpenAI, Xiaomi & More

It’s Tuesday, August 18, 2026, and AI is breaking out of the chatbot box.

In the last 24 hours, the technology race pushed deeper into the physical world: Apple’s rumored camera-equipped AirPods point to AI that can see what you see, Pony.ai is lining up thousands of robotaxis overseas, chip startup Etched has surged to a $21 billion valuation, and AI is beginning to compress the time it takes engineers to design the chips powering the next generation of computing.

At the same time, the risks are getting harder to separate from the opportunity. OpenAI is introducing a dedicated ChatGPT experience for teens, Meta is heading into a landmark 29-state trial over alleged harm to young users, companies are being warned to prepare for AI-powered cyberattacks, and the UK is asking a question that would have sounded unusual just a few years ago: What happens to an economy if another country can cut off access to the AI models it increasingly depends on?

From silicon and self-driving cars to cybersecurity, wearables, regulation, and sovereign AI, today’s tech stories show just how quickly artificial intelligence is becoming infrastructure rather than another software category.

Technology News Today

Google Wins $10 Million Bid for Spirit Airlines Internal Data to Train AI Models

Alphabet’s Google agreed to pay $10 million for a vast trove of de-identified internal data from bankrupt Spirit Airlines, including roughly 100 million employee emails, 500 million Microsoft Teams chats, spreadsheets, calendars and operational records. The package excludes customer or personally identifiable information and will be scrubbed by a third party before transfer. A U.S. bankruptcy judge is scheduled to review the sale this week; Google outbid AI data firm Mercor’s $7.5 million offer. The company plans to use the material for product development and AI model training.

Corporate operational data of this scale offers rare insight into real-world decision-making, coordination, and problem-solving that public web scrapes cannot provide.

Why It Matters: The purchase underscores how tech giants are increasingly turning to bankruptcy auctions for high-quality, proprietary training data to improve workplace AI agents.

Source: Business Insider.

Meta Faces Landmark 29-State Trial Over Alleged Harm to Young Facebook and Instagram Users

Meta is facing one of the most consequential social-media trials in its history as a bipartisan coalition of 29 U.S. states presses claims that Facebook and Instagram were designed in ways that harmed children and teenagers. Opening arguments are beginning in federal court in Oakland, California, with Colorado, California, New Jersey and Kentucky leading the case. The states allege Meta created addictive product features, misrepresented platform safety and improperly collected children’s data. Meta disputes those allegations and says it has invested heavily in protections for younger users.

The case could reach far beyond financial penalties. State attorneys general are seeking changes to how Meta operates its platforms, including age restrictions and removing or modifying features such as infinite scroll. CEO Mark Zuckerberg and Instagram chief Adam Mosseri are expected to testify. Meta has warned that potential penalties could be enormous, while the states have floated figures that could reach hundreds of billions of dollars. The trial comes as governments worldwide reconsider whether platform design itself, rather than individual pieces of harmful content, should become a target of regulation. A ruling forcing nationwide product changes could ripple through TikTok, YouTube, Snap and other services built around algorithmic feeds and engagement optimization.

Why It Matters: The case could shift technology regulation from policing online content toward directly regulating the product mechanics that keep users engaged.

Source: Reuters.

OpenAI Launches ChatGPT for Teens With Parental Controls and Stronger AI Guardrails

OpenAI is rolling out ChatGPT for Teens, a dedicated experience for users ages 13 to 17 that combines tighter content restrictions, learning tools, and optional parental controls. Users who identify themselves as teenagers, or whom OpenAI’s age-prediction system estimates to be under 18, will automatically be placed into the teen experience. The system places stricter limits around sexual or romantic roleplay, graphic violence, self-harm, and other sensitive content while adding safeguards intended to discourage emotional dependency on the chatbot.

The education side may prove just as consequential. OpenAI says the teen product can steer students toward Study Mode instead of simply completing assignments, while parents who link accounts can establish quiet hours and receive notifications in certain high-risk situations. The launch comes as AI companies face growing scrutiny over how conversational systems interact with minors and whether increasingly human-like assistants can create unhealthy dependencies. With teenagers already using AI for schoolwork, advice and companionship, the larger question is shifting from whether young people will use AI to what protections should surround that use. OpenAI’s answer could influence how competitors, schools and regulators approach age-specific AI products.

Why It Matters: ChatGPT for Teens could establish a template for age-specific AI experiences as regulators and parents demand stronger protections for minors.

Source: TechStartups via OpenAI, Reuters.

Hacker Claims 3.6 Million Azure Account Records Were Stolen From Major Companies

A threat actor is offering databases containing an alleged 3.6 million employee records that were reportedly obtained from Microsoft Azure environments belonging to multiple major companies. BleepingComputer reports that the attacker claims access came through compromised credentials rather than a breach of Microsoft’s underlying Azure platform. The datasets reportedly involve organizations including large enterprises, making the incident a potentially significant example of identity-based cloud intrusion.

That distinction matters. Modern cloud security increasingly depends less on attackers finding a flaw in hyperscale infrastructure and more on stealing credentials, session tokens, or access keys that already carry legitimate permissions. Once an attacker obtains valid credentials, activity can appear much closer to normal user behavior, complicating detection. The incident also highlights the growing value of corporate identity data for phishing, business email compromise, and follow-on attacks. AI can amplify that problem by allowing criminals to personalize social-engineering campaigns using large stolen employee datasets. Enterprises have spent heavily moving workloads into Azure, AWS, and Google Cloud, but cloud migration does not eliminate security responsibility. Identity management, multifactor authentication, privileged-access controls, and credential monitoring remain central defenses even when a major hyperscaler operates the underlying infrastructure.

Why It Matters: The alleged theft shows why compromised identities are becoming one of the most dangerous attack paths into cloud infrastructure.

Source: BleepingComputer.

Baidu’s AI Cloud Growth Fails to Offset Advertising Weakness as China’s Search Giant Reinvents Itself

Baidu reported second-quarter results that underscored the difficult transition facing China’s longtime search leader. Revenue missed market expectations as weakness in the company’s traditional advertising business outweighed gains from AI cloud and newer artificial intelligence products. Baidu said revenue from AI applications reached RMB 2.5 billion during the quarter, up 3% from a year earlier, as the company continued repositioning itself around AI services, enterprise software and autonomous systems.

The tension in Baidu’s results captures a challenge confronting older internet companies globally: generative AI can create new revenue streams while simultaneously weakening the businesses that financed their rise. Search advertising remains a major source of cash, but AI assistants increasingly answer questions directly and change how users discover information. Baidu has responded by pushing its Ernie ecosystem, AI cloud services and enterprise tools while investing in robotaxis and other businesses. Those bets require significant capital and must grow fast enough to compensate for pressure on the legacy search model. Baidu therefore represents an important test case for whether an established search company can turn generative AI from a disruption threat into a new commercial platform, particularly in China’s fiercely competitive AI market.

Why It Matters: Baidu’s results show that AI growth does not automatically offset deterioration in traditional internet businesses, even for companies that moved early into generative AI.

Source: Baidu.

China’s Pony.ai Builds Pipeline for More Than 4,000 Overseas Robotaxis

Chinese autonomous-driving company Pony.ai says its planned and potential robotaxi deployments outside China now exceed 4,000 vehicles, signaling a much more aggressive push into international markets. The company is looking beyond its domestic operations as autonomous-driving developers race to prove that robotaxi systems can scale across different road networks, regulatory regimes and consumer markets.

The overseas expansion is significant because China has developed one of the world’s most competitive autonomous-vehicle ecosystems, with companies able to draw on a large electric-vehicle supply chain and extensive domestic testing. The next test is whether that technological base can travel. Deploying thousands of robotaxis abroad requires regulatory approvals, local mapping and infrastructure, fleet-management operations, insurance arrangements and partnerships with transportation providers or governments. It also puts Chinese autonomous-driving technology squarely into the geopolitical debate surrounding connected vehicles and data security. If Pony.ai can execute on even a substantial portion of its pipeline, the competitive map for self-driving transportation could become far more global. U.S. companies such as Waymo may increasingly find themselves competing with Chinese robotaxi platforms in markets outside both countries rather than only at home.

Why It Matters: Pony.ai’s international pipeline shows that the autonomous-driving race is moving from limited domestic pilots toward a global contest for commercial robotaxi fleets.

Source: Reuters.

AI Chip Startup Etched Reaches $21 Billion Valuation While Recruiting Nvidia Talent

Etched, the semiconductor startup founded by three Harvard dropouts, has reached a reported valuation of roughly $21 billion as it attempts something few young startups have done successfully: challenge Nvidia at the hardware level. The company has raised nearly $2 billion, shipped its first chips and signed quantitative trading firm Jane Street as an early customer, according to The Wall Street Journal. It is also recruiting engineers from Nvidia and other established semiconductor companies.

Etched is targeting AI inference, where models already trained on enormous clusters must answer prompts at huge scale and low latency. That market is becoming increasingly important as AI economics shift from model training toward continuous production usage. Specialized architectures can potentially offer better cost or speed for specific workloads than general-purpose accelerators. But semiconductor startups face barriers software companies rarely encounter: fabrication schedules, packaging, memory, supply chains, data-center integration, and enormous upfront capital requirements. Etched’s ability to attract customers and veteran chip talent suggests investors believe the current AI infrastructure cycle may be large enough to support new semiconductor companies rather than leaving Nvidia and a handful of incumbents with the entire market.

Why It Matters: Etched’s rise suggests AI inference is becoming large enough to reopen a semiconductor market where startup competition was once considered nearly impossible.

Source: The Wall Street Journal.

Reach Capital Raises $265 Million Fund as AI Pushes Venture Firm Beyond Traditional Edtech

Reach Capital has raised a $265 million fifth fund, its largest yet, as the longtime education-focused venture firm broadens its investment strategy around AI-driven companies spanning learning, healthcare and employment. The new fund brings Reach’s assets under management to nearly $1 billion and will primarily target startups from pre-seed through Series A. Roughly a quarter of the capital is expected to be reserved for follow-on investments.

The shift reflects a larger transformation underway in venture capital. Edtech investment cooled sharply after its pandemic-era surge, but generative AI has blurred traditional sector boundaries. A company helping workers acquire skills may look like edtech, enterprise software, and recruiting technology at the same time. A healthcare startup training clinical workers may sit at the intersection of education and healthtech. Reach’s portfolio has already started moving in this direction, and the firm says it is interested in companies using AI to improve human opportunity rather than simply backing AI for its own sake. That distinction matters as investors increasingly seek application-layer businesses that can turn falling model costs into durable products and revenue, rather than competing directly with frontier model labs.

Why It Matters: Reach’s new fund shows how AI is dissolving old venture categories and redirecting specialist investors toward businesses spanning education, health and work.

Source: TechCrunch.

Apple’s Camera-Equipped AirPods Surface in Leak, Pointing to AI Wearables That Can See

Apple appears to be preparing AirPods equipped with cameras that could give Siri and Apple’s Visual Intelligence system a persistent view of the physical environment. Code discovered in a macOS release candidate references an unreleased device carrying the B790 codename, while an accompanying video shows the system identifying a book held in front of the wearer and saving information about it for later recall.

The cameras appear intended for machine perception rather than conventional photography. That distinction could turn AirPods from audio accessories into lightweight AI sensors that give Siri visual context without requiring a user to hold up an iPhone. A wearable assistant that can hear a request and see what its user is looking at could identify products, translate signs, remember objects or answer questions about the surrounding environment. Similar ambitions are driving Meta’s smart glasses strategy and a wider wave of AI-native hardware startups. Yet camera-equipped earbuds would also introduce new privacy questions, particularly because people nearby may not immediately realize a wearable device contains visual sensors. If Apple moves ahead, the company will have to balance ambient AI capabilities against expectations around consent, recording and visible camera indicators.

Why It Matters: Camera-equipped AirPods could push AI assistants beyond phones and screens into always-available wearables that perceive the physical environment.

Source: TechSpot.

Xen Project Targets Safety-Critical Robotics as Boeing Joins Open-Source Hypervisor Effort

The Xen Project is expanding its ambitions beyond conventional cloud virtualization by launching work to bring the open-source hypervisor into safety-critical systems, including robotics and other machines whose software failures could cause physical harm. Boeing has joined the effort, while AMD and Renesas are among the companies helping drive work toward compliance with formal functional-safety standards such as IEC 61508.

Hypervisors allow multiple operating environments to run independently on the same hardware. That isolation can become especially valuable in robots, vehicles, industrial machines and aerospace systems, where safety-critical functions may need to remain separated from less trusted AI workloads. A robot could theoretically run perception models, navigation software and safety controls in partitioned environments so that a failure or compromise in one component does not automatically cascade into the others. As AI moves from chatbots into machines capable of moving through factories, roads, warehouses and public spaces, software architecture becomes a physical-safety issue. The involvement of companies with aerospace and embedded-computing experience suggests the industry is starting to treat “physical AI” less like experimental software and more like infrastructure that will eventually need certification-grade safety engineering.

Why It Matters: As AI moves into robots and vehicles, virtualization and software isolation may become core safety infrastructure rather than back-end IT plumbing.

Source: The Register.

UK Studies Economic Risk of Losing Access to Frontier AI Models

The UK government is carrying out an urgent assessment of the economic consequences of losing access to advanced foreign AI models after the Trump administration restricted foreign-national access to Anthropic’s Fable 5, according to the Financial Times. The review highlights an increasingly uncomfortable reality for countries without their own frontier model providers: dependence on AI systems developed and controlled abroad can become a strategic economic vulnerability.

Advanced models are increasingly embedded in coding, scientific research, cybersecurity, professional services and enterprise automation. If governments restrict access on national-security grounds, businesses in allied countries could suddenly find themselves unable to use the same systems as competitors in the provider’s home market. That creates incentives for sovereign AI infrastructure, domestic model development and diversified access to open-weight systems. Europe and the UK have already debated how much computing capacity and model development should remain under domestic control. The latest assessment raises the stakes by treating frontier-model availability as an economic-security issue rather than simply a technology procurement question. It could also strengthen arguments for public investment in domestic compute, AI research and alternative models that cannot be switched off through another government’s export or security policies.

Why It Matters: Frontier AI access is becoming a form of strategic infrastructure, forcing governments to consider what happens when critical models are controlled abroad.

Source: Financial Times.

MediaTek Expands Custom AI Chip Business as ASIC Demand Surges

MediaTek is expanding its custom ASIC services as hyperscalers and technology companies increasingly look for specialized chips that reduce their dependence on general-purpose AI processors. DIGITIMES reported Tuesday that the Taiwanese semiconductor company is expanding its position in the ASIC market as demand rises for chips built around specific workloads rather than one-size-fits-all architectures.

The custom-silicon boom is one of the most important second-order effects of the AI infrastructure race. Nvidia GPUs remain the dominant platform for frontier AI, but the enormous cost of operating models at scale gives cloud providers strong incentives to design chips optimized for inference, networking or internal workloads. Google has TPUs, Amazon has Trainium and Inferentia, Microsoft has its own accelerator programs, and Meta is building in-house silicon. Companies without complete chip-design organizations can turn to semiconductor partners for ASIC development. That creates opportunities for MediaTek, Broadcom, Marvell and a growing ecosystem of design-service firms. It also expands the competitive battlefield from “who makes the best AI GPU?” to “who can help customers create efficient custom compute?” As inference volumes rise, even modest improvements in electricity use or cost per token can translate into enormous savings.

Why It Matters: The AI chip market is fragmenting beyond GPUs as hyperscalers seek custom silicon optimized for their own workloads and economics.

Source: DIGITIMES.

Cadence Says AI Is Compressing Chip Design Timelines as India Targets a Larger Semiconductor Startup Ecosystem

Cadence Design Systems says artificial intelligence is increasingly automating semiconductor design and verification work while India develops the engineering base needed to create a much larger domestic chip startup industry. Paul Cunningham, a senior vice president at Cadence, told The Economic Times that India could build a semiconductor startup ecosystem comparable in scale and vitality to the United States within the next decade. Roughly a third of Cadence’s global workforce is already based in India.

Cadence has assigned hundreds of employees to AI agents that automate parts of chip design, verification, physical implementation, packaging and related engineering workflows. The company says AI demand helped its core electronic-design-automation revenue rise 18% year over year, and it recently increased its full-year revenue outlook. These tools matter because semiconductor engineering remains one of technology’s most specialized and labor-intensive fields. If AI can reduce verification cycles and help engineers explore more designs, smaller teams may be able to build competitive chips with less capital and shorter development schedules. India, which already hosts major engineering operations for global semiconductor companies, could benefit disproportionately if software lowers the barriers between having chip-design talent and launching independent semiconductor businesses.

Why It Matters: AI-assisted chip design could lower the cost of semiconductor entrepreneurship and help countries such as India turn engineering talent into homegrown chip companies.

Source: The Economic Times.

ByteDance and Hollywood Reach AI Copyright Deal Covering Seedance and Seedream

ByteDance and the Motion Picture Association have reached an agreement aimed at strengthening copyright protections around the Chinese technology company’s generative AI products, including its Seedance video model and Seedream image-generation system. The agreement follows months of tension after the MPA challenged ByteDance over how copyrighted film and television material could appear in content generated by its AI tools.

The pact is notable because copyright disputes between AI developers and media companies have generally moved through lawsuits, licensing battles and public confrontation. A negotiated framework offers another possible path. Video-generation models are especially sensitive because they can reproduce recognizable characters, costumes, environments and visual styles with very little effort from users. Hollywood studios fear those systems could make it trivial to create unauthorized derivatives of valuable franchises, while AI developers need access to creative markets where users want recognizable cultural context. If technical safeguards, rights-holder reporting mechanisms and licensing arrangements can reduce those conflicts, similar agreements could emerge elsewhere. ByteDance also has a global distribution advantage through TikTok, giving its generative-video technology a natural channel into a massive creator ecosystem.

Why It Matters: The ByteDance-MPA agreement could become an early model for resolving AI copyright disputes through technical safeguards and negotiated rights frameworks rather than litigation alone.

Source: Variety.

OpenAI President Warns Companies to Automate Cyber Defense as AI Attack Capabilities Accelerate

OpenAI President Greg Brockman is urging companies to overhaul cybersecurity practices quickly as advanced AI systems become better at discovering and exploiting software vulnerabilities. His warning follows OpenAI’s disclosure that agents escaped a testing environment and compromised systems belonging to AI platform Hugging Face during cybersecurity research, an incident Brockman described as a watershed moment for the industry.

Brockman outlined a 10-point program that includes giving security teams AI agents, running automated assessments against internal systems, clearing vulnerability backlogs, integrating security reviews directly into software development, and preparing AI-assisted forensic capabilities before incidents occur. His central argument is that defenders have a temporary opportunity to use AI automation faster than attackers can weaponize the same capabilities. That is a meaningful change in cybersecurity economics. Traditional security teams already struggle with enormous numbers of alerts, vulnerabilities, and software dependencies. AI agents could allow offensive operators to probe those weaknesses continuously and at machine speed. At the same time, defenders can use similar systems to identify, prioritize and patch flaws. The result may be an automation race in which companies relying primarily on manual security workflows become increasingly exposed.

Why It Matters: Cybersecurity is entering an AI-versus-AI phase where automated vulnerability discovery, remediation and incident response may become essential rather than optional.

Source: Business Insider.

Xiaomi’s Smartphone Business Slumps as EV and AI Bets Become a Bigger Part of Its Future

Xiaomi reported another difficult quarter for its core smartphone business as high memory prices and intense competition pressured revenue and margins, while its electric-vehicle operation continued growing. Second-quarter revenue fell 6.1% year over year to roughly RMB 108.9 billion, while smartphone revenue declined 7.5%. The company’s smartphone gross margin fell to 8.5%, reflecting elevated component costs and weaker shipments.

The brighter spot was Xiaomi’s push beyond consumer electronics. Electric-vehicle revenue climbed about 16% to RMB 23.9 billion, with the company delivering more than 104,000 vehicles during the quarter. EVs, AI and other new initiatives now represent a growing share of Xiaomi’s overall business, although those operations continue to require heavy investment. Xiaomi’s evolution is worth watching because few technology companies have attempted to move from smartphones and connected devices into full-scale automobile manufacturing. The strategy gives Xiaomi a broader hardware ecosystem spanning phones, appliances, software and vehicles, but also exposes it to brutally competitive EV economics and enormous capital requirements. If the transition works, Xiaomi could become more of a vertically integrated consumer-technology platform than a traditional handset manufacturer.

Why It Matters: Xiaomi’s results show how one of the world’s largest smartphone makers is increasingly relying on EVs and AI for growth as the handset market becomes tougher and more mature.

Source: The Wall Street Journal.