Fireworks AI suffers second security breach in October

Key Takeaways

• Fireworks AI suffered a second security breach in October, exposing billing data, account metadata, and incomplete model weights for tens of thousands of files; the company is backed by Nvidia and AMD and valued at $17.5 billion.
• Pillar, an AI security startup, agreed to a $150 million acquisition by Vanta, bringing its SAIL framework that maps 91 risks across the AI lifecycle.
• Cyera launched endpoint security tools and specialized AI agents including Cy Detect and Cy Invest to protect corporate data used by AI systems.
• Fortinet and Equinix partnered on a Distributed AI Hub with FortiAIGate, a security system that inspects AI traffic using GPUs without slowing workloads.
• Memories.ai recorded over 1 million hours of robot training video in its first month and aims for 10 million hours by March 2027, using a sub-45-gram wearable camera.
• GenRobot AI raised $100 million from investors including Baidu and SAIC Motor to build robot training data infrastructure with its Matrix software platform.
• OpenAI and Upwork found AI-related freelance postings became 45% more diverse between 2023 and 2025, with larger budgets but similar hourly rates, and 43 categories lost market share where ChatGPT use was strongest.
• David Ellison, heading merged Paramount and Warner Bros., called AI the biggest creative breakthrough since CGI and said a streaming merger could happen within a year using AI tools.
• Anthropic started its IPO roadshow with a valuation between $1.1 trillion and $1.3 trillion, while Nvidia considers another $10 billion investment ahead of the listing.
• The European AI Act takes effect in 2026, requiring companies to document how AI contributed to B2B sales communications and to use accurate, representative data in AI systems.

Fireworks AI reveals new breach after cloud provider hack

Fireworks AI disclosed a new security breach on October 23 after a third-party cloud provider was hacked. Attackers accessed development machines and stole API keys, billing information, and some incomplete model weights. The company says customer payment card numbers were not at risk. Fireworks is working with Palo Alto Networks Unit 42 to investigate and secure its systems.

Fireworks AI notifies customers of second hack in weeks

Fireworks AI told customers on Monday about a second security breach in a few weeks. The company has over 10,000 customers including Notion, Cursor, and Harvey. The breach came through an external cloud provider and affected tens of thousands of files with customer data. Fireworks is backed by Nvidia and AMD and was valued at $17.5 billion in July.

Fireworks AI discloses second security breach details

Attackers entered Fireworks AI through a third-party cloud provider that hosted research machines. The breach exposed account metadata, billing information, and for some accounts, stored secrets and incomplete model weights. Fireworks says its Inference, Training SDK, and Nexus products were not affected. The company advises customers to rotate API keys and stored secrets.

Fireworks AI breach widens to expose billing data

Fireworks AI revealed on Monday that its earlier security investigation found more malicious activity. The breach now affects account and billing information along with secrets, datasets, and incomplete models. The company disclosed the original incident in early October. The full scope of customer impact is still being investigated.

Fireworks AI suffers second attack with wider data exposure

Fireworks AI disclosed another cyberattack in October involving billing records, datasets, and incomplete model weights. The company suspects financially motivated criminals linked to LLMJacking. Affected data was grouped into resource metadata, billing information, and model weights. Fireworks worked with Palo Alto Networks Unit 42 to fix the attack by rotating credentials and isolating systems. The company did not say how many customers were affected.

Memories.ai pays gig workers to film robot training videos

Memories.ai, a startup founded by ex-Meta researchers Shawn Shen and Ben Zhou, pays people to record first-person video of everyday tasks for robot training. Its LUCI Pin camera weighs under 45 grams and attaches magnetically to the body. The company launched Memories Upload, a gig platform where contributors record tasks like cooking or repairing equipment. The company recorded over 1 million hours of video in its first month and aims to collect 10 million hours by March 2027. It has raised $16 million in venture funding and uses Nvidia Cosmos tools to label and organize the footage.

GenRobot AI raises $100 million for robot data tools

GenRobot AI, a Beijing-based company, raised $100 million in the third quarter of 2026 to expand its robot training data infrastructure. Investors included SAIC Motor, Baidu, Eastworlds, Shoucheng Holdings, Galaxy General Robotics, GL Ventures, CITIC Goldstone, CCB International and Huatai Zijin. The company builds recording systems that capture video along with human hand and body movements. It also develops a software platform called Matrix that synchronizes, cleans and labels the recorded data. The funding reflects growing interest in collecting human demonstrations to train robots for tasks in homes, factories and logistics facilities.

Upwork AI gigs cost more and require broader skills

A joint study by OpenAI and Upwork found that AI-related freelance work is becoming more diverse and more expensive. AI-related postings became 45% more diverse between early 2023 and late 2025, growing from about 15 tasks to about 21. These projects carried larger fixed-price budgets and more hours, though hourly rates stayed similar. Of 63 work categories analyzed, 43 lost market share on Upwork where ChatGPT use was strongest. OpenAI said ChatGPT use alone does not explain the demand shifts and called the gap the capability overhang.

OpenAI and Upwork study shows AI shifting freelance demand

OpenAI and Upwork published research showing that freelance work is changing in the AI era. Some services are losing market share while AI-related projects combine different skills and involve larger budgets. The study matched ChatGPT use from January 2025 to June 2026 against Upwork job postings. Clients began requesting AI in many kinds of work, including chatbot development, video production and marketing automation. One example brief asked for AI-generated video, editing and motion graphics in a single delivery. The research also noted that freelance knowledge work generated $1.5 trillion in U.S. earnings in 2024.

David Ellison praises AI as creative breakthrough for Hollywood

David Ellison, CEO of Skydance and head of the merged Paramount and Warner Bros. studios, called AI the biggest creative unlock since CGI went mainstream. He said AI will help studios make content faster and cheaper. Ellison noted his empire has 40 franchises worth over $1 billion and serves 203 million-plus subscribers. He compared AI's potential impact to the advent of CGI in the early 1990s that enabled films like Jurassic Park and Terminator 2. He also said the merger of Paramount+ and HBO Max within a year is possible because of AI tools. Skydance is seeking $6 billion in cost-synergies from the merger including AI workflow improvements.

Fortinet and Equinix launch secure AI hub with FortiAIGate

Fortinet partnered with Equinix to build the Distributed AI Hub, which connects AI agents, models, data, and cloud providers securely. FortiAIGate acts as a security guardrail that inspects AI traffic, enforces policies, and protects sensitive data. The system uses GPUs to handle security checks without slowing down AI workloads. FortiGate VM on Equinix Network Edge also supports network security. Leaders said the approach helps enterprises secure AI across distributed environments.

European AI Act reshapes automated B2B sales rules

The European AI Act will have major rules on documentation and transparency take effect in 2026. Companies that use AI in B2B sales must clearly show that AI helped generate communication sent to prospects. Tools like Vanelo AI that automate lead generation need to keep records of how prospects were selected. The law also requires AI systems to use accurate and representative data. Companies that prepare now can build trust with prospects and avoid compliance problems.

AI security startup Pillar agrees to $150M deal with Vanta

Pillar, an AI security startup based in San Francisco, agreed to be acquired by Vanta for about $150 million in cash and stock. Pillar was founded in late 2023 and had raised just over $9 million before the deal. Its technology helps find hidden AI systems, test their weaknesses, and watch their behavior. Pillar also created the SAIL framework, which maps 91 risks across seven stages of the AI lifecycle and has been downloaded over 50,000 times. Vanta said the acquisition will help it offer active AI security beyond compliance.

Cyera launches agent-based security tools for AI era

Cyera introduced Cyera Endpoint and Cy Agents to expand its security platform for AI-driven environments. Cy Agents adds specialized agents like Cy Detect, Cy Triage, Cy Learn, Cy Investigate, and Cy Assist to help teams find and fix risks faster. Cyera Endpoint adds an on-device sensor that brings AI security and data loss prevention to corporate laptops. The tools aim to give security teams visibility into how employees and AI agents use sensitive data while keeping people in control of final decisions.

Tips to avoid bad AI-generated gifts this holiday season

This article warns shoppers about low-quality AI-generated products, often called AI slop, when buying gifts. It suggests checking for repeated product listings, identical item photos, and signs of AI artwork like blurry details or misspelled text. Shopping in person at local stores can reduce the risk of getting fake or poor-quality items. The article also warns against relying too much on AI shopping assistants and says to watch for unnatural videos or suspicious product descriptions.

Anthropic starts roadshow before Thanksgiving IPO

Anthropic plans to begin its roadshow this week ahead of a potential IPO next week before Thanksgiving. The AI company has not set a final price yet, but investors value it between $1.1 trillion and $1.3 trillion. Anthropic relies heavily on Nvidia chips, with the chipmaker investing up to $10 billion last November and considering another $10 billion for the IPO. The timing is critical as rival OpenAI, valued at $1.25 trillion, may go public soon, making this a competitive moment for AI investments.

Sources

• Fireworks AI discloses new breach after cloud provider hack
• Nvidia-Backed Fireworks AI Notifies Customers of Its Second Hack in Weeks
• Fireworks AI Discloses Second Security Breach
• Fireworks AI Breach Widens to Expose Customer Billing Data and Incomplete Model Weights
• Fireworks AI gets burned twice as customer data exposure widens
• Exclusive: In the race for robot training data, ex-Meta researchers’ startup turns to gig workers with cameras
• GenRobot AI raises $100 million to expand robot training data infrastructure
• Upwork AI gigs cost more, mix more skills
• Work at the Frontier: AI’s effect on freelance work
• David Ellison Is Very Bullish On AI’s Creative Potential
• Fortinet collaborates with Equinix on AI security
• How the European AI Act transforms B2B sales
• Months after setting its sights on Miami, AI security startup Pillar lands $150M acquisition deal with Vanta
• Cyera Expands Its Platform to Build Out a New Security Architecture for the Agentic Era
• 4 Tips To Help Avoid Buying Awful AI Slop Gifts This Holiday Season
• Anthropic Can Now Start Its Roadshow. The Traffic’s Bad.