Noam Shazeer Jumps to OpenAI, Robotaxis Hit Recall Potholes, and CRISPR’s Long Road to Everyday Medicine — The Week in Tech

AI’s Talent War Just Got Personal

Noam Shazeer’s Move from Gemini to OpenAI

The biggest AI hiring story of the month broke on June 18th: Noam Shazeer, one of the co-leads of Google DeepMind’s flagship Gemini model, has joined OpenAI. Shazeer isn’t just another researcher—he spent two decades at Google, left in 2021 to co-found Character.AI, and was lured back in 2024 as part of a reported $2.7 billion acquisition that brought his team back into Google’s orbit. Now, barely a year later, he’s crossing the aisle to OpenAI. That kind of odyssey signals how intensely competitive the frontier-model race has become.

For Google, losing a Gemini co-lead is a meaningful morale and credibility blow. Gemini has been Google’s answer to GPT-4-class models, and Shazeer was instrumental in scaling its reasoning and long-context capabilities. For OpenAI, landing a veteran of this caliber helps close the talent gap with rivals and validates its push to keep building the most capable general-purpose models. Watch for news about which product group he joins—reasoning alignment, safety, or infrastructure—because that will hint at OpenAI’s near-term priorities.

Anthropic’s Export-Control Headache

While the AI world digested Shazeer’s defection, Anthropic was fighting a very different battle in Washington. On the evening of June 13th, Anthropic received a U.S. export-control directive requiring it to suspend access to its new Claude Mythos 5 and safeguarded Fable 5 models for any foreign national, including Anthropic employees. The order reportedly gave the company a 90-minute ultimatum to comply—an almost impossible timeline that CEO Dario Amodei met by immediately engaging Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross.

The immediate trigger was a report suggesting that a jailbreak could bypass Fable 5’s safety guardrails. Anthropic pushed back, saying the jailbreak was narrow and that comparable capability exists in other public models, including OpenAI’s GPT-5.5. The broader subtext, according to reports, concerns sensitive access to the model by a group with suspected Chinese links—something that could permanently alter how frontier AI labs distribute advanced systems globally.

The Autonomous-Vehicle Industry Is Growing Up—the Hard Way

Waymo Recalls Thousands of Robotaxis

Autonomous driving had its own dose of reality this week. Waymo announced a recall of nearly 3,900 robotaxis after incidents in Arizona and San Francisco showed its vehicles failing to recognize lane closures at highway construction zones. According to a filing with the NHTSA, vehicles had driven past ramp-closure signs into restricted lanes—a serious safety lapse that forced Waymo to pause its freeway service in affected regions. The company says a software update will resolve the issue, but the recall highlights a pattern: autonomous systems trained on dense urban data still struggle with the chaos of roadwork infrastructure.

Uber Expands Robotaxi Ambitions to Houston

While Waymo paused routes, Uber was busy mapping its next expansion. Having already targeted San Francisco for a late-2026 launch of Lucid-built, Nuro-powered robotaxis, Uber is now gearing up for Houston. The company has secured a 50,000-square-foot maintenance facility and charging infrastructure, signaling real capital commitment rather than vaporware announcements. Houston will put Uber’s autonomous stack against Waymo and Tesla in a crowded Texas arena.

Rivian’s R2 Launch, Then Layoffs

Not all EV news was happy. Rivian—fresh off launching its much-anticipated R2 SUV—cut less than 2% of its workforce within days, according to a Wall Street Journal report. The company confirmed the move as a restructuring to support its first sustained path to profitability. Rivian’s pattern of layoffs (2022, 2023, 2024, 2025, and now 2026) mirrors the broader electric-vehicle sector’s transition from growth-at-all-costs to operational discipline. Legacy OEMs are feeling the pressure too: Ford’s rumored sub-$30,000 EV—possibly smaller than the Maverick—shows how price competition is reshaping the mass-market segment.

Biotech’s Quiet Revolution: The CRISPR Rollout Continues

From Lab Breakthrough to Patient Bedside

AI and cars grabbed headlines, but a quieter transformation is underway in biotech. The rollout of CRISPR-based gene therapies—after years of hype-hope-scandal cycles—is entering a more pragmatic phase. Companies like Vertex and CRISPR Therapeutics have already won regulatory approval for sickle-cell and beta-thalassemia treatments, and the next generation of candidates is tackling more common conditions: hemophilia, certain cancers, and even cardiovascular risk factors.

What’s notable now is not just scientific progress but the operational reality of scaling these therapies. Manufacturing gene therapies at industrial volumes remains a bottleneck; each treatment is essentially a personalized bioprocess. Quality control, viral-vector supply chains, and reimbursement negotiations with insurers are now the limiting factors, not the underlying science.

ADCs and the Pharma Gold Rush

On the drug-development front, antibody-drug conjugates (ADCs) are the current obsession. These therapies combine monoclonal antibodies with cytotoxic payloads, delivering toxins directly to cancer cells. Daiichi Sankyo—whose ADC expertise helped ignite the category—is now racing to stay ahead of competitors piling into the space. The result is a crowded, expensive, but genuinely innovative pipeline that could reshape oncology over the next five years.

The Long Wait for Accessible Biotech

Even as clinical results improve, the economics of novel therapies remain formidable. A single CRISPR treatment can cost hundreds of thousands to over two million dollars per patient, which poses a stark equity problem. Progress is undeniable—efficacy is real, safety is better understood, and regulatory pathways are clearer—but translating that into affordable, widely available medicine will require market-design innovations as much as scientific ones.

Putting It All Together: Three Threads, One Theme

The AI, automotive, and biotech stories of this week share a common lesson. In AI, talent mobility and government export controls are becoming as decisive as technical breakthroughs; the frontier is not just a scientific boundary but a geopolitical one. In autonomous vehicles, the gap between demonstration and safe, scalable deployment keeps surfacing in very visible ways—construction-zone confusion today, regulatory scrutiny tomorrow. In biotech, the chasm between laboratory proof and accessible therapy is filled not with science but with manufacturing, reimbursement, and equity.

What connects all three is the transition from potential to deployment. Phase transitions are always messier than they look from a distance, and this week reminded us that real-world complexity—politics, profit motives, flawed sensors, and sparse infrastructure—does not yield to clever models or strong laboratory results alone. The teams that navigate these frictions will define the next phase of the decade, not the ones that simply announced the most impressive demos.