AI Safety stories
Businesses could slash AI spending as Glean says its desktop workspace used 81% fewer tokens than Claude Cowork in tests.
Errors buried in workflows, reports and design decisions are the real danger, as AI mistakes become harder to spot and trace.
Enterprises weighing autonomous AI controls now have a former HSBC security chief on Token's board as it targets risky agent actions.
Enterprises face rising exposure to AI-specific attacks as spending on tools to secure models and apps nears USD $4.8 billion in 2027.
Banks can now use Google's new AI tools for regulated research and risk work, with audit trails and controlled data access in preview.
Rogue behaviour in testing has heightened fears that autonomous tools are being handed sensitive systems without proper oversight or limits.
Compliance with the EU AI Act will see Claude text tagged worldwide at launch, as Anthropic says it cannot yet limit watermarking by region.
AI rollouts are pushing SRE and platform teams to monitor models, manage incidents and fill governance gaps in production.
Australian firms can speed month-end processing from days to an hour by running Kiraa AI on Macs instead of costly cloud servers.
The beta tool lets Claude users trade prediction markets and pay other agents without leaving their coding session, though security risks remain.
The rollout will let law firms and other clients run AI inside governed repositories without weakening confidentiality or audit controls.
Security gaps are slowing business adoption of AI agents, as most technology leaders warn that governance and access controls lag behind their autonomy.
It is aimed at firms that want AI agents to update records and trigger workflows without losing oversight or audit trails.
Demand for engineers who can link AI to industrial systems is rising as manufacturers race to automate safely and stay competitive.
Sales teams will soon be able to query live Salesforce data inside Claude as the pair deepen an AI tie-up across CRM and Slack.
High-risk AI obligations have been pushed back, but transparency, GPAI and banned-practice rules remain in force for many firms.
More deals and faster lender integrations are expected as Lendi shifts its mortgage platform towards agentic AI for 1,300 brokers.
Governance gaps are slowing enterprise AI adoption, with 79% of Australian DevSecOps professionals saying compliance is getting harder.
Australian enterprises risk extra compliance burdens as AI-made observability tools shift governance obligations onto in-house teams.
Rising scam losses in Australia are pushing businesses to verify texts, images and voice notes before paying or acting on urgent requests.
.webp)
