The U.S. Department of Justice has indicted two engineers from Volkswagen on charges of securities fraud due to an alleged insider-trading scheme related to the automaker’s partnership with Rivian.

The indictment, made public on Friday by the U.S. Attorney for the Southern District of New York, claims that Michael Stamp and Marcus Plank profited over $300,000 by leveraging confidential insider knowledge. They are accused of purchasing Rivian stock and options after discovering that Volkswagen and the EV manufacturer intended to establish a joint venture — referred to internally as “Project Climb” — but before any announcements were made to the public.

Rivian and Volkswagen revealed their plans for the joint venture on June 25, 2024, aimed at developing electric vehicle architecture and software. Volkswagen originally pledged to invest $5 billion in Rivian, with funds being available as the companies meet specific goals. The venture has since expanded to $5.8 billion, with Volkswagen now holding the position of Rivian’s largest shareholder.

Following the announcement in June, Rivian’s stock experienced a 23% increase. Stamp and Plank reportedly offloaded their investments in Rivian, with Stamp earning around $250,000 in profits, Plank generating approximately $50,000, and a close family member of Plank making about $12,000, as outlined in the indictment.

U.S. Attorney Jay Clayton stated in a press release on Friday:

Michael Stamp and Marcus Plank’s purported misuse of their employer’s confidential information enabled them to garner over $300,000 in illicit gains. When individuals exploit confidential information for personal financial benefit, they impair the foundational principles that ensure our markets operate fairly and effectively. Insider trading is a crime that the residents of New York expect to be prosecuted vigorously. Its ramifications permeate the financial ecosystem, affecting regular investors and diminishing public trust. Today’s indictment highlights our Office’s and our law enforcement partners’ dedication to safeguarding the integrity of our markets and holding accountable those who opt to breach the law.

Investigators assert that the two engineers were aware that their actions were unlawful. Eight days before the joint venture was publicly disclosed, Stamp searched for “statute of limitations insider trading,” and a close relative of Plank searched, in German, “how is insider trading prosecuted?,” according to the indictment.

The two, both residing in San Jose, were apprehended on Friday and are set to appear in the U.S. District Court for the Northern District of California. The proceedings have been assigned to U.S. District Judge Katherine Polk Failla. Stamp and Plank could face up to 25 years in prison if found guilty of federal securities fraud.

Rivian has opted not to comment. A spokesperson from Volkswagen stated that the company is aware of the Department of Justice’s actions today concerning two individuals.

“The proceedings pertain to specific individuals and do not involve allegations against the company,” the Volkswagen spokesperson communicated in an email statement. “As this matter is ongoing, we are unable to provide additional comments.”

The article was revised to include Volkswagen’s response.

Poke, the AI companion you chat with like a buddy, is gearing up for its next significant advancement. The organization behind this assistant, The Interaction Company of California, has been acquired by AI coding startup Cognition in a transaction that values the startup at “low nine figures.”

This agreement will integrate Poke’s interaction style and character into Cognition’s coding assistant Devin. In turn, Poke will leverage Cognition’s frameworks and models to enhance its speed and dependability, according to Cognition.

What draws consumers — and now its purchaser — to Poke is its way of interacting with user inquiries. Rather than acting merely as a tool, Poke communicates in a friendly manner, responding as a friend and even weaving in slang and humor. This could bring added value to Devin’s coding capabilities.

“You likely prefer co-workers with personality over those that feel robotic,” Interaction Company co-founder Marvin von Hagen mentioned in a TechCrunch interview, where he also disclosed the acquisition’s value. “When your co-workers are software engineers, they can share jokes, and that proves enjoyable and entertaining.”

The acquisition highlights a rising conviction that the manner in which AI assistants interact with users will become as significant as the foundational models that drive them. Cognition aims to transform Devin from mere software into a true colleague.

“The Interaction team has developed an agent that users adore: it’s proactive, it understands you, and it’s enjoyable to converse with. This is precisely how working with Devin should feel, and now we can build it collaboratively,” stated Cognition co-founder Scott Wu in a blog entry.

He also pointed out that both he and Cognition co-founder Walden Yan had invested as angel investors in the company that developed Poke.

First released in March 2026, Poke enables users to interact with its AI agent via their preferred messaging platform, whether iMessage, SMS, Telegram, or WhatsApp in certain regions. Users turn to Poke for a range of tasks across areas such as travel, health, finance, scheduling, and learning.

Von Hagen highlighted that productivity tasks such as managing emails, reminders, and to-do lists rank among Poke’s most frequent applications.

According to Cognition, Poke users sent over 100 million messages through the platform in the last three months. However, despite serving hundreds of thousands of users, Poke had been costly to operate, which made profitability challenging, von Hagen noted.

In June, Poke achieved a milestone by becoming the first AI agent authorized to operate on Apple’s Messages for Business platform, which provides a standardized method for businesses to manage customer communications within Apple’s Messages app.

Looking ahead, no immediate changes are expected for Poke until the year’s end, as stated. Poke intends to continue on Apple’s platform.

However, in the coming year, the team will explore how both Poke and Devin can advance, with Poke utilizing Cognition’s latest software engineering model SWE-1.7 for certain tasks. The complete integration of both products remains a possibility as well.

“I believe in the long run that Poke can enhance reliability in coordinating all these coding [tasks]… and concurrently, Devin could evolve to be more like Poke,” von Hagen remarked. “At present, Devin can only handle one pull request at a time… I see considerable value in allowing Poke to manage various Devin sessions.” Moreover, Poke could assist Devin in recalling tasks across different sessions.

“Having a consistent co-worker would be beneficial,” he concluded.

The U.S. Justice Department is taking legal action against an American accused of giving U.S. border officials a passcode that erased the data on his phone, as stated in an indictment and media coverage. 

This case is believed to be the first instance in the U.S. in which federal prosecutors have charged an individual for purportedly destroying data using a “duress” password embedded in the phone’s operating system.

According to The Guardian, which reported on the matter after the initial court hearing on Monday, Atlanta resident Samuel Tunick is challenging the allegations. Tunick’s legal representatives argue that it was illegal for U.S. Customs and Border Protection to confiscate his phone when he returned to the U.S. last year and that any evidence — including the supposed erasure of his phone — should be dismissed.

The case revolves around a function found in GrapheneOS, a tailored Android operating system that replaces the software typically on most modern Google Pixel devices. Tunick’s legal team confirmed that GrapheneOS was operational on his phone.

This software feature allows the phone’s owner to configure a passcode that intentionally erases the device’s data if entered instead of the standard unlock passcode.

Tunick’s situation also brings to light ongoing debates about the constitutional rights that can be invoked at the border, which the U.S. government has historically maintained is not considered U.S. territory until an individual is granted permission to enter.

The government’s indictment, which contains a typo (“Untied States Code”), claims Tunick allegedly provided a passcode to border officials that led to the phone “erasing the digital contents” before the device was taken.

Tunick’s legal team submitted a motion to suppress the evidence, asserting that the detention and confiscation were unlawful. The motion detailed that U.S. border authorities had subjected Tunick to a secondary inspection at Atlanta’s Hartsfield-Jackson airport upon his return from abroad on January 24, 2025, but he was continuously denied access to legal counsel and was not informed of his rights.

Tunick’s lawyers alleged that the government insisted on accessing his phone under the guise of searching for child exploitation material, yet failed to provide justification for its suspicions. His motion to suppress contended that the government was actually investigating him due to his affiliation with an enduring environmental movement named Defend the Atlanta Forest, which opposes the establishment of a vast training facility for law enforcement in Atlanta known as “Cop City.”

The motion stated that the border agents claimed they did not require a warrant to probe Tunick’s phone since he had not yet entered U.S. territory. The U.S. government has consistently asserted that it can inspect and seize individuals’ devices without a warrant or court authorization until they are allowed entry into the United States.

When Tunick provided his passcode and authorities entered it, “the screen went blank, flashed several times, and the phone seemed to restart.” The authorities confiscated his phone anyway, informing him that he was free to leave and could enter the United States.

Prosecutors subsequently charged Tunick under a federal law that renders it illegal to intentionally destroy or damage property to obstruct authorities from seizing it. Tunick has pleaded not guilty.

Matthew Dodge, an assistant federal public defender on Tunick’s legal team, informed TechCrunch that it is exceptionally uncommon to see this federal law applied in an indictment.

Security professionals also indicated that they had not encountered charges filed in this manner before.

Bill Budington, a senior staff technologist at the Electronic Frontier Foundation, and Runa Sandvik, a digital security expert who aims to protect vulnerable individuals as the founder of the security consultancy firm Granitt, told TechCrunch that they had not encountered analogous cases involving the application of duress passwords.

“I have not seen this before, although I’ve discussed potential scenarios with activists and journalists throughout the years,” remarked Sandvik. “I believe this case serves as a reminder that authorities may claim you intentionally destroyed data, so it’s advisable to not carry that data when crossing certain boundaries.” 

“With some advanced planning, you can always download the necessary data once you reach your destination,” added Sandvik. 

The Electronic Frontier Foundation offers resources on how to safeguard your data and security at the U.S. border, detailing your rights.

The Atlanta federal court presiding over the matter is anticipated to rule on Tunick’s motion to suppress later this year. A spokesperson for the Justice Department declined to provide comments when contacted by TechCrunch.

Updated to correct the spelling of a surname in the fifteenth paragraph.

Anduril, a defense technology firm, is reportedly seeking to secure a fresh round of funding that could elevate its valuation significantly by $40 billion, reaching approximately $100 billion, according to reports from Reuters citing unnamed sources.

The organization, which gathered $5 billion in May at a valuation of $61 billion — nearly double the $30.5 billion value it achieved in its Series G round back in June 2025 — might be arranging this fundraising as a two-phase endeavor, with the latter phase involving investors at an escalated valuation. Both phases are expected to wrap up within the current year, as per Reuters.

This development arrives amidst an extraordinary resurgence in the defense tech sector, driven by a surge in demand for drones, autonomous vehicles, and the incorporation of AI in military operations, particularly in the ongoing conflicts in the Middle East and Eastern Europe. In the first half of this year, venture capital investments in the industry surged to over $12 billion, more than doubling the nearly $10 billion raised by startups throughout 2025.

Anduril has capitalized on this increasing demand, securing contracts with the U.S. Department of Defense, Air Force, and Army; the Dutch Ministry of Defence; NATO; the U.K. Ministry of Defence; Poland; and additional entities.

The company reported in May that it had more than doubled its revenue to $2.2 billion in 2025 compared to the previous year.

Other startups are also benefiting: Military aircraft producer Shield AI raised $1.5 billion in March, Mach Industries saw its valuation rise fourfold to $1.8 billion last month, and Europe’s Helsing secured $1.8 billion this month at an $18 billion valuation.

A common theme among these transactions is a pivot towards affordable, more expendable hardware, often referred to as “attritable” systems, departing from the costly-to-replace equipment that has characterized defense contracting for many years. Mach CEO Ethan Thornton has indicated that the startup is developing systems suited for today’s warfare at significantly lower costs than conventional defense contractors, using Ukraine’s deployment of autonomous drones as a benchmark.

In addition, both Mach and Anduril have taken steps to secure their own propulsion supplies. Mach purchased solid rocket motor manufacturer Exquadrum for $50 million in May, while the Pentagon has supported Anduril’s initiatives to enhance domestic production capacity for solid rocket motors — efforts addressing a supply chain issue that has existed before and is associated with the trend for low-cost weapons.

Anduril’s backers comprise Thrive Capital, Andreessen Horowitz, Founders Fund, ICONIQ, Flux Capital, Greycroft, Altimeter, 1789 Capital, and current U.S. Vice President JD Vance, as reported by PitchBook.

“No decisions have been made regarding any future financing, and anyone claiming to know the terms, structure, pricing, or timing is either speculating or misinformed. As a private entity, we consistently assess opportunities to fund business growth,” Anduril stated in an emailed response.

Note: This story was updated to incorporate Anduril’s statement.

For several months, major AI companies have implemented specially approved programs and stringent safeguards to restrict the use of their models by malicious actors. However, these restrictions are now impairing the efforts of legitimate network defenders as well as offensive cybersecurity researchers. 

In June, the U.S. government imposed export control measures on Anthropic’s widely discussed AI models, Mythos and Fable. This action was partly triggered by a report suggesting that it might be feasible to circumvent the models’ safeguards designed to stop users from employing them to create and carry out harmful cyberattacks.

Regardless of whether the situation was genuinely influenced by concerns regarding a jailbreak, it’s evident that Anthropic has frequently promoted Mythos as a doomsday cyber tool that can only be entrusted to thoroughly vetted users, and even then with rigorous safeguards enforced. (The export restrictions on Fable 5 and Mythos 5 have since been rescinded. Fable 5 returned to public access on July 1; Mythos 5 has been reintegrated solely to screened U.S. organizations as part of the government’s evaluation process.)

This type of gatekeeping is not exclusive to Mythos. Both Anthropic with its other models and OpenAI offer cybersecurity researchers avenues to apply for vetting, and — if approved — gain access to models with reduced cybersecurity limitations: OpenAI’s Trusted Access for Cyber initiative and Anthropic’s Cyber Verification Program. 

These safeguards have faced significant criticism, especially from researchers tasked with uncovering unknown vulnerabilities within systems and formulating ways to exploit them before malicious actors do.

In a recent appearance on a cybersecurity podcast, Mark Dowd, a renowned security researcher, expressed that “it’s not really comfortable to me that these random large companies are making arbitrary decisions about what is safe in security and what’s not.”

Dowd has spent years identifying and selling “zero-days” — previously unidentified software flaws and the exploits that leverage them — to Western governments instead of reporting them to the software developers for patching. Governments pay a premium for vulnerabilities because they remain unaddressed, beneficial for intelligence activities.

Dowd acknowledged that his work might color his perspective, but he is not the only one. Multiple individuals working in offensive cybersecurity — who actively probe systems for weaknesses — described to TechCrunch their use of AI tools and how they navigate their guardrails. 

Chris Anley, the chief scientist at security consulting powerhouse NCC Group, mentioned that prompting an AI model to attempt to exploit a bug is crucial in verifying that it’s a genuine vulnerability and worth addressing. However, if a guardrail leads the model to outright refuse to respond, it adversely affects defenders, he stated.

“This is where the entire offensive versus defensive and guardrails aspect comes in, because ‘fix this code’ as a prompt serves as both a vital mechanism for defense and a blueprint for identifying critical vulnerabilities within the code base,” Anley explained. “Thus, the same tool functions as both an offensive and a defensive tool, and the two cannot actually be separated.”

“It’s ‘like a hammer,’” he continued. “You cannot construct a house without a hammer. It is undoubtedly a tool but irreducibly also a weapon.”

When he and his team encounter such hurdles, they sometimes revert to open-source AI models that come with no safeguards whatsoever.

Paolo Stagno, the chief technology officer at Crowdfense, a well-known entity that develops, procures, and sells undiscovered vulnerabilities to government bodies, shared Dowd’s sentiments, stating that AI companies “essentially treat clients like children who require supervision” with their vetted programs and guardrails. 

Stagno stated that he and his team do utilize frontier models — but strictly for reverse engineering. They refrain from employing AI to assist in identifying vulnerabilities or constructing exploits, as integrating that work into a cloud-based model poses risks of exposing sensitive vulnerability information or having it absorbed into future training sessions. For that purpose, he stated they resort to locally run open-source models, avoiding data sharing outside the model. 

Giuseppe Cali, a security researcher specializing in zero-days and exploit development, claimed that guardrails do not obstruct his work. This is because he does not use AI for offensive tasks; rather, he employs it for initial reverse engineering, to comprehend the code he’s analyzing, and to create supporting tools. For that, he emphasized that AI tools can expedite the process and allow him to concentrate on discovering vulnerabilities. 

“I still want to control the actual bug discovery and weaponization myself, and that wouldn’t change if all guardrails were removed tomorrow,” Cali remarked. “I am possessive about my bugs, and I enjoy this game too much to allow models to play it for me.”

One researcher at a smartphone-component manufacturer, who requested anonymity due to lack of authorization to speak to the media, indicated that his employer is not part of Anthropic’s CVP program, leaving its tools nearly ineffective for discovering vulnerabilities due to overly strict guardrails.

“If it catches wind we’re doing anything security related, it just stops and isn’t usable,” the person remarked. 

Chris Thompson — chief executive of cybersecurity firm RemoteThreat and founder of Offensive AI Con, an event focused on offensive security and AI — stated that in his experience with frontier AI models, the guardrails can be erratic and vary in performance daily. This inconsistency holds true even within the more lenient frameworks of Anthropic’s and OpenAI’s vetted programs. 

“I think the practical effect is that you spend a lot of time negotiating with the model instead of concentrating on the fundamental security program,” Thompson remarked. “Rather than analyzing a vulnerability and reasoning through the exploitability, you are attempting to uncover why you are encountering inconsistent results or why models are over-sanitizing the output.” 

Consequently, researchers are driven towards or rely on Chinese open-source models like GLM — models that can be freely downloaded and executed locally without any vetting or usage constraints — noted Thompson.

“You have these responsible researchers that are being pushed away from U.S.-governed systems to foreign-owned systems,” he remarked. “I believe it’s more damaging than beneficial to have these guardrails in place.”

Instead of imposing further restrictions, Thompson urged the AI frontier labs to expand their programs, provide responsible access, and hold those who misuse their tools accountable. Otherwise, he contended, defenders will lose the AI race.

“There’s this big storm approaching. There’s this significant wave of attacks that are going to occur at unprecedented speed and scale,” Thompson warned. “But the same security consulting companies and legitimate researchers striving to make a difference are currently being stifled.”

TechCrunch Startup Battlefield is heading to Australia — and we’ve teamed up with Stripe to discover the nation’s most thrilling early-stage startups. On August 19, eight Australian startups will present live at Stripe Tour Sydney in front of investors, media, and the tech community. Three of them will be awarded prizes. One will secure an automatic entry to TechCrunch Disrupt 2026, occurring in San Francisco from October 13-15.

What’s the key to identifying the next standout founder? The judges.

Startup Battlefield has a proven history of success. The competition has launched more than 1,700 companies worldwide, including Dropbox, Cloudflare, Discord, and Trello. Alumni have collectively raised $32 billion and achieved over 250 exits. During the last Startup Battlefield in Sydney, HealthMatch emerged as the winner — now valued at over $25 million with a million patients globally.

Such success isn’t coincidental. It stems from having the right individuals present, those who can identify potential, pose challenging questions, and create opportunities for founders. This year, we’ve gathered three judges representing the pinnacle of Australia’s startup landscape and global fintech innovation.

Meet your judges

Jasmine Liew, Head of Startup & Investor Partnerships for APAC, Stripe

Stripe not only sponsors Startup Battlefield — the company is also intricately woven into the startup landscape. Jasmine Liew possesses institutional insight into what it entails to scale on a global level and the payment infrastructure that underpins contemporary commerce. When a founder presents before her, they are pitching to someone who comprehends not just the vision, but also the intricacies of transforming that vision into a profitable venture.

Brendan Hill, angel investor and venture partner, Ten13

Brendan Hill embodies the investor segment that actively provides funding. As a venture partner and engaged angel investor, he not only assesses ideas — he’s also considering whether he would invest his own capital in a team. That perspective is crucial when founders seek to secure their subsequent round.

Jarron Aizen, founder and CEO, Hapana

Jarron Aizen is a founder himself, bringing his understanding of the experience these eight pitchers will face. He’s traversed the journey from concept to company, and that insight is incredibly valuable. When he poses a question from the judge’s panel, it is backed by the triumphs and challenges of developing a business in the Australian landscape.

What’s at stake

The grand prize champion will receive $15,000 in Stripe fee credits — tangible funds that can expedite growth — along with automatic entry into Startup Battlefield 200 at TechCrunch Disrupt 2026. This is not merely a reward. It’s a launchpad. The second place winner will get $5,000 in Stripe credits, and the third place winner will receive $2,000 in Stripe credits. More significantly, all finalists will have the opportunity to interact with these judges, gain media exposure, and connect with the Stripe Tour Sydney audience comprised of founders, operators, and investors. 

The true spectacle occurs on August 19

If you’re an investor, a founder not ready to pitch, or simply someone keen to witness the evolution of Australian startups, Stripe Tour Sydney on August 19 is the event to attend. Register now.

The stage is ready. The judges await. The question remains: Which startup will shine next?

Hosted and MC’d by Isabelle Johannessen, director of Startup Battlefield.

Chip manufacturer AMD is setting its sights on rival Nvidia with its newest hardware introduction: a rack-scale solution crafted to meet the computing demands of the globe’s largest AI laboratories.

At the company’s packed Advancing AI conference in San Francisco on Thursday, AMD Chair and CEO Dr. Lisa Su showcased the new AI rack system named Helios — along with its increasing roster of clients, which includes Microsoft — as the firm gears up for its release later this year. Su also promoted the firm’s latest chips designed to satisfy the high demands of the AI sector.

Rack systems merge multiple processors into one robust unit. They are specifically designed for data centers, where they facilitate the training and operation of AI models and other computation-heavy tasks.

Su described Helios as the industry’s “top-performing AI rack,” noting that it was “engineered to train and execute the most demanding frontier models globally at an unprecedented scale.” The system is set to be implemented by prominent AI enterprises at gigawatt-scale, according to the company.

Historically, Nvidia has led this field with its Vera Rubin and Grace Blackwell rack-scale systems. AMD is clearly aiming to get a piece of the pie. Helios’ performance indicators seem to offer it a tangible opportunity, reportedly surpassing Vera Rubin in various metrics, as noted by The Register.

Helios, unveiled in 2025 and presented live in January at CES 2026, already counts some notable clients among its base, such as OpenAI, Meta, Oracle, Anthropic, and Microsoft, all of which intend to utilize the system. Microsoft CEO Satya Nadella mentioned on Monday that the company plans to enhance its Azure infrastructure with Helios. Additionally, Anthropic and AMD declared a strategic alliance on Wednesday to deploy up to two gigawatts of GPUs through the new rack system.

On Thursday, AMD also unveiled its Venice-X CPU, engineered for data centers and capable of managing high-computing workloads. The Venice-X is projected for launch in 2027.

During her speech, Su reflected on the chip industry’s path, asserting that by 2030, chips supporting AI will constitute a significant portion of the overall computing market. This transformation is propelled by a “step change in compute demand,” largely fueled by the emergence of agentic AI, she stated.

“When you instruct the agent to perform a task, it goes through numerous steps, reasoning, calling tools, accessing data, and repeating the process until it resolves the issue; thus, a multitude of GPUs is required to manage all of that,” the executive explained.

“We anticipate that by 2030, the AI accelerator market will reach approximately $1.4 trillion,” Su stated. “This indicates that by decade’s end, the AI accelerator market will come close to matching the total size of the current semiconductor market.”

“We do foresee that GPUs will represent the bulk of that market since the algorithms are still in their early stages, and we’re observing continual shifts in workloads, which favors programmability within the complete silicon ecosystem,” she added.