Predictions about network security tend to age badly, but a few directions are already visible in how organisations buy and operate it. Security functions are being consumed as services instead of installed as products. Machine traffic is growing faster than human traffic. Encryption standards are facing a scheduled replacement. Each of these pushes work towards a cloud-delivered control layer, and together they explain why converged architecture is becoming the default rather than an alternative.
Current Adoption of Cloud-Delivered Security Inspection
Adoption has moved past early experiments. Most large organisations now run some portion of their inspection in the cloud, usually starting with remote access and web traffic, and the practical questions have shifted from whether the model works to how far it should extend. Descriptions of how SASE services are delivered now cover branch connectivity, private application access, SaaS control and workload traffic under one policy, which is considerably broader than the remote-access replacement many deployments began as.
That breadth creates a new dependency. When connectivity and security arrive from the same provider, availability engineering and contract terms become security concerns, and resilience planning has to account for the control layer itself rather than only the systems behind it.
Machine Identities, API Calls and Non-Human Traffic
The traffic mix is changing. Service accounts, automation pipelines, integration jobs and increasingly autonomous software agents now initiate a large share of connections, and none of them authenticate the way a person does. Access decisions built entirely around user identity and device posture cover a shrinking fraction of real activity.
This is where application programming interfaces become a network security problem rather than a development one. The widely used ranking of critical API security risks places broken object level authorisation first, describing endpoints that accept object identifiers as a broad attack surface for access control flaws. A control layer sitting between workloads can enforce which service may call which endpoint, but it cannot repair authorisation logic inside the application. Expect provider roadmaps to concentrate here, and expect the governance work to land with application owners rather than network teams. The inventory problem comes first in either case, because most organisations cannot currently list the service accounts and integrations operating inside their own estate.
Post-Quantum Cryptography Migration and Transport Security
Every inspected session depends on certificates and key exchange, which makes the move to quantum-resistant algorithms an architectural event rather than a maintenance task. Government programmes have shifted from general advice to dated commitments: European member states and the Commission published a coordinated post-quantum transition roadmap in June 2025, setting out a synchronised migration path rather than leaving each country to improvise. Several national authorities have published their own staged timelines alongside it, and the common thread is that discovery alone is expected to take large organisations years rather than months.
A cloud-delivered inspection layer helps here in a specific way. Algorithm support is updated by the provider at the edge rather than by upgrading hardware at every site, which shortens the part of the migration involving transport security. It does not solve the harder problems: finding where certificates and keys live, replacing embedded cryptography in applications, and sequencing changes so systems keep interoperating.
Policy as Code: Version-Controlled Security Configuration
Configuration is moving towards version-controlled definitions reviewed like software changes. The reasons are mundane but durable: audit evidence is easier to produce, mistakes are easier to reverse, and a policy that exists as text can be tested before it reaches production. Teams adopting this early tend to find their entitlement data is the limiting factor, not the tooling. In practice, a rule change then arrives as a reviewed commit rather than an untracked console edit nobody can account for later.
Data Residency and Where Decryption Is Permitted
Regulatory pressure increasingly determines where inspection may happen. If decryption occurs outside a permitted jurisdiction, a technically sound deployment becomes a compliance problem. This is already reshaping provider build-outs and buyer requirements, and it favours architectures where enforcement location can be pinned per policy rather than assigned by whichever edge location is nearest. Worth asking providers how that pinning behaves during a regional failure, since the fallback path is where residency commitments quietly break.
Provider vs Customer Responsibility in SASE Services
| Area | Handled by the provider | Remains with the customer |
| Transport cryptography | Algorithm support at the edge | Keys and certificates in applications |
| Inspection capacity | Scaling and edge build-out | Sizing against real peak traffic |
| Policy enforcement | Consistent application of rules | Deciding entitlements and scope |
| Machine identities | Service-to-service enforcement | Authorisation logic inside the code |
| Logging | Collection and retention | Export, correlation and review |
| Availability | Platform and regional failover | Continuity planning around outages |
What Will Separate SASE Services Providers
Feature lists are converging, so the differences worth testing are operational:
- Breadth of edge locations relative to where your users and regulated data sit
- Whether inspection for a given policy can be constrained to specific regions
- Quality of telemetry export, including whether raw logs can leave the platform
- Transparency during incidents affecting the provider’s own infrastructure
- Support for workload and machine identities, not only user accounts
How to Prepare for a SASE Migration
Several pieces of preparation hold their value whichever direction the providers end up taking:
- Build an accurate inventory of applications, service accounts and automated integrations
- Clean up entitlement data before expressing it as granular policy
- Start cryptographic discovery now, since it is slow and nobody else can do it for you
- Keep policy definitions exportable so a provider change remains possible
- Treat the control layer as critical infrastructure in continuity planning
None of this work depends on a purchase decision, which is rather the point. Each item is work an organisation would benefit from having done anyway, and each one removes a dependency that would otherwise slow a migration down.
Conclusion
The direction of travel is less dramatic than vendor messaging suggests. Network security is becoming a subscription with centralised policy, broader coverage of machine traffic, and an explicit cryptographic deadline attached. None of that removes the organisational work: deciding entitlements, maintaining inventories, and owning the risk that comes with concentrating connectivity and inspection in one place. Organisations that do the unglamorous preparation get the benefit of the architecture; the rest inherit the same problems in a newer form.
Frequently Asked Questions
Will cloud-delivered security replace on-premises firewalls entirely?
Unlikely in the near term. Data centres, industrial sites and environments with strict latency or availability requirements continue to justify local enforcement, managed under the same policy where possible.
How do autonomous software agents change access control?
They act with delegated permissions and no human present, so entitlements need scoping and expiry at the agent level. Treating them as ordinary user accounts grants far broader reach than intended.
Do SASE services help with post-quantum readiness?
Partly. Transport-level algorithm support becomes the provider’s responsibility, which removes site-by-site hardware work. Discovery of keys and certificates and changes inside applications remain internal tasks.
What should organisations avoid committing to in a SASE contract?
Architectures that cannot be unwound: proprietary policy formats with no export path, logging that never leaves the platform, and contracts that bundle connectivity and security with no separable renewal. Click here see more details.
