Gemini Enterprise Agent Platform
Google Cloud has launched a comprehensive Enterprise Agent platform powered by Gemini, designed as a central hub for businesses to construct, expand, govern,
and refine autonomous AI agents. This platform offers flexibility, allowing technical teams to select existing models or develop their own bespoke AI agents. Among the advanced models available are Gemini 3.1 Pro, Gemini 3.1 Flash Image (also known as Nano Banana 2), and Lyria 3, alongside offerings from Anthropic like Claude Opus 4.7 and its Sonnet and Haiku variants. The platform is enhanced with integrated features for agent deployment, DevOps practices, orchestration, and security, ensuring that employees can seamlessly deploy agents through the Gemini Enterprise app while maintaining strict control, governance, and security at scale. This integrated approach contrasts with fragmented service offerings, providing a vertically optimized stack where all components are co-developed for the efficiency and scale required in the current era of production AI. The technology is presented as a foundational element for businesses to build their growth engines.
Next-Gen TPU Innovations
Google has also revealed its eighth generation of custom Tensor Processor Units (TPUs), comprising two specialized architectures: TPU 8t for training and TPU 8i for inference. These chips are engineered to facilitate the creation of highly capable AI models, manage complex AI agent swarms, and perform intricate reasoning tasks. The TPU 8t is optimized for training large AI models within weeks, balancing high compute throughput, shared memory, and interchip bandwidth with superior power efficiency and productive compute time. In contrast, the TPU 8i features 288 GB of high-bandwidth memory and 384 MB of on-chip SRAM, which is triple the capacity of its predecessor. This increased memory allows for faster agent responses, enabling tasks to be delegated and completed in under five seconds by utilizing on-chip memory exclusively. Both the TPU 8t and TPU 8i are slated for general availability later this year, positioning them as a competitive alternative to existing GPU solutions in the AI hardware market.
AI-Powered Cybersecurity Agents
To bolster business defenses against increasingly sophisticated AI-driven cyber threats, Google Cloud has introduced a suite of new AI-powered security agents. The Dark Web Intelligence agent leverages Gemini models and the expertise of the Google Threat Intelligence Group to construct security profiles for organizations by analyzing millions of daily dark web events with remarkable accuracy, pinpointing only relevant threats. Complementing this, the Threat Hunting Agent utilizes Google's internal threat intelligence to enable security teams to proactively identify novel attack patterns and adversary behaviors that elude traditional security measures. Furthermore, the Detection Engineering Agent automates the generation of persistent detection rules for various threat scenarios, transforming a historically manual process into an efficient, automated workflow. In addition to these agents, reCAPTCHA is being integrated into Google Cloud Fraud Defense to help distinguish between legitimate users, bots, and unauthorized agents, thereby combating malicious bot activity and enhancing the security of digital commerce transactions from account creation to payment.
