Cisco has reported that cybercriminals have been taking advantage of a flaw in one of its widely-used networking solutions for large corporations for a minimum of three years, leading the U.S. government and its allies to recommend organizations to implement protective measures.

This flaw, which has been rated with a maximum vulnerability severity score of 10.0, enables cybercriminals to remotely infiltrate networks utilizing its Catalyst SD-WAN systems, which assist large businesses and governmental bodies with multiple locations in connecting their private networks across extensive distances.

By leveraging this vulnerability over the internet, hackers can obtain the highest level of access to these devices and sustain covert ongoing access within a target’s network, permitting them to monitor or extract data over an extended timeframe.

Cisco indicated that after identifying the flaw, its analysts uncovered evidence of exploitation dating back to 2023. Some organizations affected are believed to be part of critical infrastructure. While the company did not disclose detailed information, “critical infrastructure” may encompass everything from electrical grids and water supply systems to the transportation industry.

Numerous governments, including those of Australia, Canada, New Zealand, the United Kingdom, and the United States, issued a notice warning that threat actors are targeting entities “globally.”

The U.S. cybersecurity agency CISA instructed all civilian federal agencies to update their systems by the end of Friday, referencing an imminent threat and unacceptable risk to the federal government. The federal cybersecurity institution, currently operating at a reduced capacity due to a partial government shutdown, acknowledged awareness of ongoing exploitation.

Neither Cisco nor the governmental agencies connected the attacks to a specific threat group or nation-state, if known, but monitored one cluster of activity designated as UAT-8616.

In December, Cisco cautioned about another similarly-rated 10.0 vulnerability in the Async software that operates the majority of its products, which was actively being exploited to penetrate its customer networks.

An assembly of prominent open-source developers is partnering with a venture capital investor to establish a nonprofit named the Open Source Endowment, aiming to provide a lasting solution to the ongoing challenge of funding open-source software development. 

Supporters of the Open Source Endowment comprise individuals such as Thomas Dohmke (the previous CEO of GitHub who secured a record-setting $60 million for his development tool startup Entire); Mitchell Hashimoto (founder of HashiCorp, sold to IBM for $6.4 billion last year); Supabase founder and CEO Paul Copplestone; one of the co-founders of NGINX; the creators of Vue.js and cURL; along with executives from Elastic, Spotify, and various others. Overall, the initiative has attracted over 50 contributors to date. 

The nonprofit, which has recently attained official 501(c)(3) recognition, has currently gathered more than $750,000 in pledges. If the vision of its founder, Konstantin Vinogradov, comes to fruition, it is expected to amass $100 million in assets in seven years. 

Vinogradov is a venture capitalist focusing on open-source, AI, and infrastructure software, having formerly served as a general partner at Runa Capital. He possesses “some experience with university endowments,” which are significant investors in venture capital funds, he told TechCrunch.  

According to Vinogradov, as he explored various open-source initiatives, a recurring issue emerged: “There is no reliable source of sustainable funding for open-source maintainers. And that’s a significant concern.” (“Maintainer” refers to developers involved in open-source projects, such as fixing bugs, selecting and verifying community-submitted features, or developing new features themselves.)

The endowment will finance projects based on criteria such as user count or the dependency of other projects on that specific software. It will also prioritize projects that lack sufficient support from grants, donations, or umbrella organizations like Linux’s Alpha-Omega. Vinogradov has already formed a board for the nonprofit.

Financially constrained and exhausted

The financial shortfall in open source is far from a novel issue. Open-source software is generally distributed at no cost, and since the community usually devotes time and labor voluntarily, up to 86% of open-source developers do not receive payment for their contributions.

This situation may not pose a challenge for hobbyists or professional developers employed by companies to maintain projects, but such a framework is fundamentally unstable. Open-source software forms the foundation of the internet, and nearly every major corporation utilizes open-source tools in some capacity. Indeed, open-source software constitutes as much as 55% of the technological architecture within organizations and is embedded in everything from databases to operating systems. 

Although it is certainly feasible for open-source developers to monetize their free projects to achieve substantial wealth, the likelihood, to paraphrase the Hunger Games, is not in their favor.  

There has always existed, and continues to exist, a dedicated group of developers who willingly devote their time and resources for free to oversee vital, influential, and essential projects. Many of these individuals are experiencing burnout.  

This concern briefly entered public awareness in 2014, during the OpenSSL Heartbleed incident, when a flaw was discovered in an open-source security project, relied upon by a majority of the internet, maintained solely by one developer. 

Numerous efforts have been made over the years to rectify the funding dilemma. Some initiatives accept contributions from corporate sponsors. For example, The Linux Foundation, which secured approximately $300 million last year mainly from corporate backers, disburses grants to specific projects through its Alpha-Omega Project. In 2025, Alpha-Omega granted $5.8 million to 14 projects, as reported.  

Certain projects receive contributions directly from corporations. In January, for instance, Anthropic donated $1.5 million to the Python Software Foundation. While the Foundation expressed gratitude for the funding, Anthropic itself raised $30 billion this month. For the AI lab, such a donation amounts to pocket change. 

Nevertheless, not every developer is inclined to accept corporate funding, as there are concerns regarding bestowing excessive influence to donor companies. For instance, there was significant controversy last year within the Ruby community regarding the departure of several long-term maintainers and its primary sponsor Spotify, as reported by The Register.  

The Open Source Endowment aims to support projects while mitigating such risks. 

“The only sustainable way to back open source is through private financing,” asserts Vinogradov. 

Why has an endowment not been attempted previously? Vinogradov explains that endowments necessitate patience. They allocate a significant portion of their assets for investment, utilizing only a small percentage of their income each year, and require years or even decades to achieve a meaningful scale.

However, if executed correctly, that patience can yield an independent fund capable of supporting critical open-source projects indefinitely.

Google has revealed today the new iteration of its well-known image creation model, Nano Banana 2. This updated model, technically termed Gemini 3.1 Flash Image, is capable of generating images that appear more lifelike than its earlier version. Furthermore, it will now serve as the default model in the Gemini app for Fast, Thinking, and Pro modes.

The initial release of Nano Banana occurred in August 2025, leading users to create countless images using the Gemini app, particularly in countries such as India. In November, the company launched Nano Banana Pro, which enabled users to generate more intricate and high-resolution images.

Nano Banana 2 keeps some of the high-detail features of the Pro model but enhances the speed of image production. The company states that users can generate images with resolutions ranging from 512px up to 4K, accommodating various aspect ratios.

With Nano Banana 2, users can ensure continuity for up to five characters and maintain the fidelity of 14 objects within a single workflow, enhancing the narrative experience. According to Google, users can also make elaborate requests with specific details for image generation. Additionally, the model allows for the creation of images with brighter lighting, enriched textures, and clearer details.

With this launch, Nano Banana 2 is set to be the standard model for image creation across all applications within the Gemini app. The company is also integrating it as the standard option for image generation in its video editing platform, Flow.

In Search, Nano Banana 2 will serve as the default for Google Search outcomes via Google Lens and in AI Mode across 141 nations via the Google app and on the web for both desktop and mobile.

For subscribers using Google’s premium plans, Google AI Pro and Ultra, the option to continue utilizing Nano Banana Pro for specialized tasks remains available by regenerating images through the three-dot menu.

For developers, Nano Banana 2 will be accessible in a preview format through the Gemini API, Gemini CLI, and the Vertex API. It will also be offered via AI Studio and the company’s development tool Antigravity, which was introduced last November.

The company announced that all images generated with the new model will feature a SynthID watermark, which signifies that they are AI-created. Furthermore, these images will be compatible with C2PA Content Credentials, a framework established by a consortium of companies including Adobe, Microsoft, Google, OpenAI, and Meta. Since the introduction of the SynthID verification in the Gemini app last November, usage has surpassed 20 million instances, according to Google.

On Thursday, a Greek court sentenced the Intellexa founder to eight years behind bars for unauthorized wiretapping and breaches of privacy, as reported by multiple sources. 

Tal Dilian, alongside three other executives from Intellexa, faced trial for involvement in a scandal referred to as “Greek Watergate,” which traces back to 2022. The Greek administration was accused of utilizing Intellexa’s spyware to intercept the communications of politicians, journalists, business figures, and military personnel.

Others sentenced today include Dilian’s business associate Sara Aleksandra Fayssal Hamou; his former deputy and Intellexa shareholder, Felix Bitzios; and Yiannis Lavranos, the owner of a company associated with Intellexa. 

Dilian did not reply to TechCrunch’s inquiry for comment. 

This marks the first documented instance of a spyware developer receiving a prison sentence due to the abuse of their technology.

In 2024, the U.S. government imposed sanctions on Intellexa along with several of its affiliated firms, Dilian, and Hamou, for their involvement in creating the spyware known as Predator, which was used to target Americans, including government officials and reporters.

The court has instructed authorities to conduct further investigations and has suspended the sentence while awaiting an appeal.

Elon Musk’s X is simplifying the process for advertisers to transfer their creative campaigns from other social media platforms to its service.

On Thursday, the organization unveiled an extended range of aspect ratio compatibility for both image and video advertisements, allowing advertisers to utilize the content they’ve developed for other platforms without needing to reformat, crop, or reconstruct their materials.

Although AI has assisted advertisers in automating the resizing of creative assets for various platforms, skipping that step entirely is more convenient for advertisers.

Thus, X now permits advertisers to upload the same assets they are using on other social media directly to its X Ads Manager through Media Studio or its Campaign Form.

This update underscores the ongoing significance of X’s advertising segment, which faced declines following the acquisition of the company, then known as Twitter, by Musk. While ad revenue saw improvement under previous CEO Linda Yaccarino, revenues in 2025 remained below levels observed before Musk’s acquisition.

According to eMarketer’s forecast from last May, X’s advertising sector was anticipated to begin revitalizing sometime last year, a prediction confirmed by Bloomberg, but would still be only half the scale it was before the sale.

X reports that the newly supported aspect ratios now include 4:5 (1440 x 1800 pixels) and 2:3 (1080 × 1620 pixels). The company already accommodates other formats such as 1:1 (1080 × 1080 pixels), 16:9 (1920 × 1080 pixels), 9:16 (1080 × 1920 pixels), and 1.91:1 (2064 × 1080 pixels).

“We are dedicated to enabling advertisers to achieve their performance objectives more effectively and impactfully,” stated Monique Pintarelli, head of global advertising at xAI, in a statement. (xAI acquired X last year.)

“With complete aspect ratio support, brands can now directly repurpose creatives on X—eliminating the need for reformatting, duplication, or compromise—while facilitating quicker testing, brand consistency, and increased reach among our highly engaged, real-time audience for superior outcomes,” she continued.