Protect AI has raised $35 million in a Series A funding round. The startup was founded by Ian Swanson, the former head of Amazon Web Services’ AI and machine learning business, and is aimed at offering a cybersecurity solution specifically for AI and ML developers. Many of the latter are now taking advantage of a growing ecosystem of tools, such as third party datasets and foundational models, and Protect AI’s “AI Radar” platform is meant to track all of these components in a “machine learning bill of materials” that can be dynamically assessed for security risks.
Foundry Technologies is reportedly in funding talks that would bring its valuation to $350 million, up from $50 million just last year, when the company did its seed round. The startup aims to rent out servers to run AI software, a model that would help other AI companies to avoid considerable hardware investments.
San Francisco-based startup Effectiv has raised $4.5 million in a new funding round, bringing its valuation over $9 million. The company has developed an anti-fraud tool designed to help financial services organizations combat the rise of AI-driven fraud attacks. Its platform analyzes user behavior in a given transaction, including what device is being used and how it’s being held (among other metrics), and scans for anomalies. An AI model trained on loan fraud patterns is compared to any detected anomalies, allowing the system to flag potential fraud in real time.
Stable Diffusion has followed through on its promise to launch a new image generation engine in July. Stable Diffusion XL (“SDXL”) 1.0 was trained on a 3.5 billion-parameter model, and is being pitched as a foundation for the development of further tools. Stable Diffusion is placing a particular emphasis on its platform’s effectiveness at ‘fine-tuning’ image generation models. The base model is now available through a range of sources, including the Amazon Bedrock and SageMaker Jumpstart platforms.
Anthropic, Google, Microsoft, and OpenAI have jointly announced an industry group focused on AI safety research and information sharing. Dubbed the “Frontier Model Forum”, the group plans to establish an advisory board, a charter, and a working group in the coming months. The move follows last week’s news that these same founding companies had agreed to a set of AI safeguards proposed by the White House.
BBC reporters got a look at Intel’s deepfake detection tool, “FakeCatcher”, and had a mixed reaction. The system, which uses photoplethysmography to detect blood flow changes, “appeared to be pretty good” at spotting deepfakes, but also flagged a number of genuine videos as deepfakes, performing more poorly when images were more pixelated. Their host, an Intel research scientist, explained that FakeCatcher was configured to be extra careful, erring on the side of caution with respect to flagging potential deepfakes.
Companies that deployed AI-driven cybersecurity defences lost 39.3 percent less money on data breaches than those that didn’t, according to an analysis of 2022 data by BanklessTimes. The organization also found that corporate networks saw a 38 percent increase in cyberattacks last year, with the United Kingdom reporting a particularly strong spike of 77 percent. Fortunately, a growing number of organizations are building defences. Just over half of those that experienced a data breach between March of 2022 and March of 2023 were planning to increase security spending.
Sumsub has announced that Singapore will serve as the home of its APAC headquarters, citing an “escalating threat of identity fraud in the region” that has coincided with the rise of a number of digital startups, including FinTech companies. Penny Chai, the former APAC Channels Director for Jumio who joined Sumsub earlier this year, will act as the firm’s leading sales representative in the region. Sumsub offers a suite of anti-fraud and KYC solutions, including a sophisticated 3D FaceMap system designed for authentication and deepfake detection. The company joined The Payments Association earlier this month.
The chatbot’s take: We asked ChatGPT to distill the most important takeaways from this week’s news. It gave us an earful:
July 28, 2023 – by Alex Perala