Fastly Launches AI Security Tools, Stock Up 14%

Fastly launched a suite of AI security tools on September 21, 2026, sending its stock up 14 percent. The new offerings include AI Firewall, AI Runtime Control, and enhanced API Security capabilities designed to give enterprises real-time control over AI model calls, token spending, and budget limits. AI Firewall evaluates prompts to block attacks such as injection attempts, while Runtime Control routes traffic and enforces spending guardrails. Cloudflare rose 7 percent and Datadog gained 5 percent in the same session as investors reacted to the announcement.

The timing matters because machine-generated traffic has been surging across Fastly's network. Automated traffic crossed 50 percent of total traffic in July and August 2026, and AI traffic grew 6.5 times faster than human traffic from January through May. Fastly's network now operates at 622 Tbps and handles over five trillion requests daily. McKinsey research found that 93 percent of organizations are exceeding their AI budgets, which makes spending controls and visibility tools increasingly important for enterprises.

In a separate development on the same day, Belgian cybersecurity firm Aikido released an open-weight AI model called Altar. The model is a compressed version of Z.AI's GLM-5.3, built specifically for cybersecurity tasks and designed to run locally so companies can keep sensitive code on their own servers. Aikido, which reached a $1 billion valuation in January 2026, plans to deploy Altar in solutions for customers including Belgian bank Belfius.

Security concerns around AI extended beyond product launches. Researchers from Hacktron broke into OpenAI's systems earlier this summer with assistance from Anthropic's Claude chatbot. They accessed private internal systems including source code after initially targeting a public messaging board. OpenAI patched the vulnerability and paid the researchers $6,500 as a bug bounty. Cybersecurity experts say the incident highlights how unprepared the AI industry is for security risks created by its own increasingly powerful technology.

NVIDIA also moved to simplify AI infrastructure with an update to its Dynamo-Triton platform. The new update allows a single application to run across multiple GPUs without complex coding, and the system automatically distributes the workload. In tests with the Cosmos 3 Nano model, using eight GPUs cut generation time from over 156 seconds to roughly 34 seconds, a significant improvement for teams processing video and media.

On the investment front, Cliff Asness sold a large position in Sandisk while buying shares in CoreWeave. Sandisk, a NAND flash supplier for AI data centers, saw its stock price jump 655 percent in 2026 and recently joined the S&P 100. Analyst Srini Pajjuri gave the stock a neutral rating despite strong quarterly earnings. Asness shifted capital to CoreWeave, betting on continued demand for AI computing power.

Treasury Secretary Scott Bessent confirmed that AI was discussed during pre-summit talks with Chinese Vice Premier He Lifeng. The conversations were formalized as USA-China AI dialogues, with plans to meet again in two months in China. Bessent said both sides want to establish a communications line for incidents and discuss protocols for managing leading AI dangers. President Trump has made outcompeting China a priority in AI.

Fed President Austan Goolsbee warned that AI spending could trigger inflation if supplies of power and construction materials run short. ETF investors are responding by targeting utilities, copper miners, oil and gas, and uranium funds to position for the physical resource demands of expanding data centers.

A McKinsey survey of over 1,700 people found a wide gap between AI productivity gains and financial returns. While 80 percent said AI improved their productivity, only 37 percent reported a positive impact on their organizations' earnings. Patrick Lafferty, who became CEO of marketing consultancy APR this summer, said companies need to redesign end-to-end workflows rather than simply deploying tools into existing processes.

Invesco's Summer 2026 survey of retirement plan participants found that investors are open to AI and private markets but want guardrails. Ninety-three percent knew private markets carry higher fees and potentially higher returns, and 65 percent expressed interest in a target date fund with a modest private markets allocation. Participants preferred AI as a copilot with human involvement over full automation, and 97 percent said the ability to set boundaries on AI-driven investment decisions was at least somewhat important.

Key Takeaways

  • <ul><li>Fastly launched AI Firewall, AI Runtime Control, and API Security tools on September 21, 2026, pushing its stock up 14 percent while Cloudflare rose 7 percent.</li><li>Machine-generated traffic exceeded 50 percent of Fastly's network traffic in July and August 2026, with AI traffic growing 6.5 times faster than human traffic from January through May.</li><li>Belgian firm Aikido released Altar, an open-weight AI model based on Z.AI's GLM-5.3, designed for local cybersecurity deployments; Aikido reached a $1 billion valuation in January 2026.</li><li>Hacktron researchers breached OpenAI's internal systems with help from Anthropic's Claude, accessing private code before OpenAI patched the issue and paid a $6,500 bug bounty.</li><li>NVIDIA's Dynamo-Triton update allows multi-GPU AI model serving without complex coding, cutting Cosmos 3 Nano generation time from over 156 seconds to about 34 seconds using eight GPUs.</li><li>Cliff Asness sold Sandisk stock after a 655 percent 2026 gain and bought CoreWeave shares, shifting capital toward AI computing demand.</li><li>Treasury Secretary Bessent confirmed formal USA-China AI dialogues following pre-summit talks, with plans to meet again in two months in China.</li><li>Fed President Goolsbee warned AI spending could cause inflation if power and construction materials become scarce, prompting ETF investors to target utilities, copper, and energy funds.</li><li>A McKinsey survey found 80 percent of respondents said AI improved their productivity, but only 37 percent reported positive earnings impact, highlighting a productivity-to-profitability gap.</li><li>Invesco's survey found 93 percent of retirement plan participants knew private markets carry higher fees, 65 percent wanted modest private markets exposure, and 97 percent wanted boundaries on AI investment decisions.</li></ul>

Fastly stock jumps 14 percent on new AI Firewall launch

Fastly launched AI Firewall, AI Runtime Control, and new API Security tools on September 21, 2026. The announcement pushed Fastly stock up 14 percent. Cloudflare rose 7 percent and Datadog rose 5 percent in the same session. Fastly reported that machine-generated traffic passed 50 percent of total traffic in July and August. AI traffic grew 6.5 times faster than human traffic from January through May.

Fastly launches AI Firewall and Runtime Control tools

Fastly introduced AI Runtime Control, AI Firewall, and new API Security capabilities for organizations. The tools give companies real-time visibility and control over their AI systems. Art Berthold, CTO of Fastly, said the products provide a secure and scalable way to manage AI. The new capabilities are part of Fastly's edge cloud platform. They are designed to meet the changing needs of customers adopting AI.

Fastly gives enterprises real-time AI model and agent control

Fastly released AI Runtime Control, AI Firewall, and API Security tools to help organizations manage AI at scale. Machine-generated traffic crossed 50 percent of Fastly's network traffic in July and August 2026. AI traffic grew 6.5 times faster than human traffic from January through May 2026. McKinsey research found that 93 percent of organizations are exceeding their AI budgets. The tools address model access, AI application protection, and agent access to enterprise APIs.

Fastly launches AI security and control tools for enterprises

Fastly announced AI Runtime Control, AI Firewall, and enhanced API Security on September 21, 2026. AI Runtime Control routes model calls and offers visibility into token spend and budget controls. AI Firewall evaluates prompts to block attacks like prompt injection. API Security enforces contracts and can block non-conforming requests. Machine-generated traffic exceeded 50 percent of Fastly's network in July and August 2026. Fastly's network operates at 622 Tbps and handles over five trillion requests daily.

Belgium's Aikido launches cybersecurity AI model Altar

Belgian cybersecurity firm Aikido released an open-weight AI model called Altar on September 21, 2026. Altar is a compressed version of Z.AI's GLM-5.3 model designed for cybersecurity use. It is available for local deployments so companies do not need to send sensitive code to external providers. The model will be used in solutions for customers like Belgian bank Belfius. Aikido reached a one billion dollar valuation in January 2026.

Belgium's Aikido releases open-weight AI model for local cybersecurity

Belgian cybersecurity firm Aikido launched an open-weight AI model called Altar for cybersecurity use. Altar is based on Z.AI's GLM-5.3 model and is designed for local deployment so companies can avoid sending sensitive code to outside providers. The model will be used in solutions for customers including Belgian bank Belfius. Aikido reached a $1 billion valuation in January.

Invesco survey shows retirement plan participants open to AI and private markets

Invesco's Summer 2026 Defined Contribution Participant Pulse Survey found that retirement plan participants are open to private markets and AI. 93% knew private markets carry higher fees and potentially higher returns, and 65% were interested in a target date fund with a modest private markets allocation. Participants preferred AI as a copilot with human involvement rather than a replacement. 97% said the ability to set boundaries on AI investment decisions was at least somewhat important.

Hackers who breached OpenAI warn AI industry lacks security readiness

Researchers from Hacktron broke into OpenAI's systems earlier this summer with help from Anthropic's Claude chatbot. They accessed private systems including internal code after first targeting a public messaging board. OpenAI patched the issue and paid the hackers $6,500 for reporting the vulnerability. Cybersecurity experts say the AI industry is unprepared for security risks created by the growing power of its own technology.

CT ad agency CEO addresses gap between AI productivity and financial returns

Patrick Lafferty became CEO of APR, a global marketing production consultancy with about 120 employees, this summer. A McKinsey survey of over 1,700 people found that while 80% said AI improved their productivity, only 37% reported a positive impact on their organizations' earnings. Lafferty said companies need to redesign end-to-end business workflows to see real returns from AI rather than just deploying tools in existing workflows.

Treasury secretary confirms AI talks with China ahead of Trump-Xi summit

Treasury Secretary Scott Bessent confirmed that AI was discussed during pre-summit talks with Chinese Vice Premier He Lifeng. The talks were formalized as USA-China AI dialogues, with plans to meet again in two months in China. Bessent said both sides want to establish a communications line for incidents and discuss protocols for leading AI dangers. President Trump has made outcompeting China a priority in AI.

NVIDIA simplifies multi-GPU AI model serving

NVIDIA released an update to its Dynamo-Triton platform that makes it easier to run generative AI models across multiple GPUs. This new feature allows a single application to use several graphics cards without complex coding. The system automatically handles the distribution of work, which significantly reduces the time needed to generate results. Testing with the Cosmos 3 Nano model showed that using eight GPUs cut the generation time from over 156 seconds to about 34 seconds. This improvement helps teams process video and media faster while keeping their software workflows stable.

Tokenized stocks and AI threaten Broadridge share

Fenimore Asset Management released a Q2 2026 investor letter warning that Broadridge Financial Solutions faces new competition. The rise of tokenized equities, which represent company ownership through digital tokens, is changing how investors buy stocks. At the same time, artificial intelligence is automating business processes, further challenging traditional financial services. These trends are causing investors to move toward digital platforms and AI-driven solutions instead of legacy systems. As a result, Broadridge may lose market share as the industry shifts toward these newer technologies.

Fed watches AI inflation as ETF investors look at utilities

Chicago Fed President Austan Goolsbee warned that AI spending could cause inflation if power and construction materials run out. He suggests that companies supplying electricity, copper, and infrastructure will become very important as AI grows. ETF investors are now looking at funds like the Utilities Select Sector SPDR Fund and the Global X Copper Miners ETF to profit from this trend. Other options include the Energy Select Sector SPDR ETF for oil and gas, and the Global X Uranium ETF for nuclear power. These investments target the physical resources needed to build the data centers that power artificial intelligence.

Cliff Asness sells Sandisk stock and buys CoreWeave

Billionaire investor Cliff Asness sold a large amount of Sandisk stock while buying shares in CoreWeave. Sandisk, a major supplier of NAND flash storage for AI data centers, saw its stock price jump 655% in 2026. The company recently joined the S&P 100 index and reported strong quarterly earnings that beat expectations. However, analyst Srini Pajjuri gave Sandisk a neutral rating, suggesting the stock might fall slightly in the near term. Meanwhile, Asness moved his money to CoreWeave, betting on the growing demand for AI computing power.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI Security Tools AI Firewall AI Runtime Control API Security Machine-Generated Traffic AI Traffic Growth Open-Weight AI Model Cybersecurity Incident AI Model Serving AI Computing Demand USA-China AI Dialogues AI Inflation Risk AI Productivity Profitability Gap AI Investment Guardrails

Comments

Loading...